Statistical Analysis of TZs

Actually I would rather keep the discussions and results on list for now. There's still some work I'm doing that isn't ready for publication yet, but apart from that I'll try to submit as much as I've discovered.

1) Sorry, I don't know much about Donchian Channels. But from what you've said, it sounds like they could be used in the First Opposite TZ system, yes.

2) Yes, the probabilty is changing all the time, espcecially in respect to where price is when you enter. You can calulate this using the indicator, at least in respect to what happened in the past. But you'll never be able to take into account everything - such as the nature of the current price action, current sentiment, trend strength, high-impact news, and so on - just by using an indicator. The probability is just a guide.

3) Yes, we call these EXs (or expirations). I haven't mentioned them yet. If your SL is miles away then they do indeed become a problem, and you have to resort to money and position management techniques to avoid a loss. It might be better to avoid EXs by using narrower stops and targets. I guess that depends on the trader.

I've never used binary options so can't comment on that. Interesting idea though, thanks.

By the way, whenever I say "we" here I'm always referring to just Jay and I. I can assure you that we aren't some mysterious institution with ulterior motives. Far from it! In fact, if anyone offers you paid coaching about TZ Systems - or any other trading system come to think of it - please walk away. It almost always means they're poor traders and want to supplement their income with easy cash from gullible newbies.

Thank you for the response NorthTrader. I am very interested to see what I can learn here and will be listening from the back from now one.
Here is a donchian Channel, its the indicator/system the turtles of the hay day used. Any one using this TZ indicator should look at this as it is literally the exact same indicator, it just appears some one, the guy who originally introduced it, sold it as something "new". h=width of the donchian channel. The turtles would use a 20 Donchian Channel on top of a 55 Donchian Channel. They traded daily charts so were basically trading 20 day and 55 day High or Low Break outs. If you lay a donchian channel over your TZ indicators, and use the same time frames or H you will see they are the exact same thing.



So by trading these overlapped zones you are really trading a Donchian Trend Reversal. if you you use H48 and the new bottom zone has formed. This is because price has made a new after a serious of highs for the specified time period, which is almost no different then the classic H,L,HH,HL, LH,LL Trend reversal. I may be wrong but these are my thoughts at a first glance of reading through everything I found on these zones.

Here is how I am thinking it could be used for Binary Options. Once a channel is touched and a new direction is taking place using H48 on an hourly chart, we know that in 48 candles there is a 90%+ probability a new low will be cleared(I would want to see how the time decay works as I think if a new low will occur it will happen in the first 1-4 candles). So a new low happens at 1.0550 and we know that in 1-4 Candles a new low will be created.

Say we sell a Call Option with a strike of 10555 that expires in 1-4 Hours. We know there is a 90%+ probability a new low will be formed below our strike price before the option expires. It does not assure us the price will close below the strike price, but it does assure us that it will go below the strike price, which we could then sell our contract early for a profit. Just some ideas. As I mentioned above, These are just my thoughts at a first glance of reviewing this TZ stuff. I will continue to sit back and listen to see what I can learn.

P.S. you could also use a bull spread of the top of the box is close enough to price, this way you do not worry about the SL or 1st TZ at all as your risk is capped and you only look to milk the trend as it continues in the direction of the last donchian channel touch.

Thanks Sis.yphus. I think Dochian channels are good for this purpose but limited by the fact that you're stuck with only one h value, for example h=20 on the chart you posted, without continuously changing the period to math the conditions. A dedicated TZ indicator like mine will show all the potential TZs within a specified range of h values (e.g. 30 to 999), so you don't need to spend time counting bars. For example, standard view:


With h-left rectangles:


First opposite TZs only (within specified range):


Regarding your binary option ideas, I'll have to chew on those some more. My past experience with conventional options wasn't very successful, so I'm a bit wary of them! But what you said makes a lot of sense. If the probability of a hit within the first few bars is high enough, and there's a good way to limit the losses on the few losers, then it sounds good.

I really appreciate your thoughts, by your way. This is one of the reasons I started this thread, to encourage smart traders like you to think about TZs from a different perspective.

I'll get back to my presentation soon hopefully. This week has been much too busy so far

Someone asked me a few posts back if my indicators were available, but I can't find it now (maybe they deleted it). Anyway, my old TZ indis are still available on the extracted threads as far as I know, or you can download them from my profile page.

The new ones, written for the topics in this thread, aren't available yet. I probably will make them available, but not just yet. They need tidying up a bit and extra documentation before being made public.

Also, all my indis are only available for NinjaTrader, sorry.

I use recurrent statistics indicator as place where price will return in a future sure. If you put order for attack this zone and price don't go before h bars, if you have balance and you don't lose that order, you can let it opened without fear. Normally are close in a relative short time. But sometimes price move much before come back. I think is better don't close it if you are sure price is in range

I'm back! Sorry for the delay. My non-trading business is unnaturally busy at the moment, so there might be lengthy gaps between posts.

So, just to recap, we'd found that there was a high probability of the first opposite PTZ being resolved, for a wide range of h values and currency pairs, especially on the M240 timeframe. However, many of these setups weren't tradeable due to very small retracements and poor R:R.

It was time to optimise the entry and SL levels based on the retracement stats, and to analyse previous setups to see if we could apply any filters to reduce the losers. It was also time to start forward testing.

Also, the naming of the TZs - 1st and 2nd - was confusing, so we started calling them Base TZ and Target TZ. The Target TZ was the first opposite PTZ following a Base TZ (PTZ or CTZ), for a given h value. The same Base TZ could be used for multiple Target TZs, with progressively higher h values up to the maximum h value of the Base TZ. A Base TZ could also be a Target TZ for the preceding Base TZ.

Confusing? Yep! So here's an example on a chart:

Here is a list of parameters (filters) we considered based on chart observations and previous stats:

h Min
TZs with h values lower than this would be ignored.
Valid values: positive integer (e.g. 30)

h Max
This was mainly to limit the processing time for cases where there were hundreds of bars of empty space to the left of the TZ bar. TZs with h values higher than this paramter would be included, but the exact h value would be unknown.
Valid values: integer >= h Min (e.g. 999)

h Base / h Target Relation
In order to match the double sequence stats posted earlier, all setups must be h Base >= h Target. Also, subsequent Target TZs must have higher h values than the previous ones, for the same Base TZ. But what about h Base < h Target setups?
Valid values: 0 (hBase < hTarget), 1 (hBase >= hTarget), 2 (All)

h Variance
This is the hardest one to explain, but basically a setup had to unique among a narrow range of smaller h values too. For example, if this was set to 2 and a previous TZ existed with h-1 or h-2, then the current TZ would be invalid.
Valid values: any integer >= 0, but typically set to 0 or 2

h Even / Odd
Didn't make much sense logically, but it seemed to make a difference for some pairs so we included it anyway.
Valid values: 0 (Evens only), 1 (Odds only), 2 (All)

h Divisible by 6
Why 6? Because 6 x M240 bars = 1 day.
Valid values: 0 (Ignore), 1 (Include)

k Min
Where k is the number of pips between the TZ tip and next closest price within h bars to the left. In other words, the height of the h-left TZ rectangle. On the chart, some of the losers had very small k values, less than double the spread of the pair. In other words, they were only just TZs. Some of the winners had small k values too, of course.
Valid values: positive integer (e.g. 5 pips)

k Min Applies To
Valid values: 0 (Base TZ), 1 (Target TZ), 2 (both)

Min Retrace
The minimum number of pips, or percentage of the distance between Target TZ and Base TZ, that price must retrace for the TZ to be considered valid.
Valid values: integer >= 0 (e.g. 20%)

Max Retrace
The maximum number of pips, or percentage of the distance between Target TZ and Base TZ, that price could reach without the setup being considered stopped out.
Valid values: integer > Retrace Min (e.g. 100%)

Recheck
If the first Target TZ within the h range is invalid due to any of the above parameter settings, then treat subsequent ones as first ones for the purpose of the report (until a valid one is found).
Valid values: 0 (off), 1 (on)

And here is a typical report that the indicator produced. Note that all parameters are set to produce as many first opposite TZs as possible for a range of h values greater than 30. This report acts as a base from which to optimise. In the report:

TP = Target TZs that are resolved before SL
SL = price breaches Max Retrace level (which in this report is Base TZ) before Target TZ is resolved
EX = price doesn't hit TP or SL within the timeframe of h-right

I'll talk about the optimisation results in my next posts.

=====================================================================
$EURUSD M240, Pip = 0.0001, Run time = 18-Apr-2015 18:10 (local)
h = 30-999+, h Variance = 0, k Min = 1 (applies to Target TZs only), Recheck = False
h Base/Target Relation = 'hBase >= hTarget' only, h Even/Odd = Both, h Div 6 = Included
Min Retrace (Entry) = 0%, Max Retrace (SL) = 100%, Min TP = 100.0%
Total bars = 9999 (from 12-Nov-2008 07:00 to 17-Apr-2015 10:00)

*** All h values here and on the chart are hMax (maximum number of left-transient bars in the TZ) ***

--------------------------------------------------
First TZs
--------------------------------------------------
Total = 189 (100.0% of all TZs)
Valid = 155 (82.0% of these TZs)

TP = 143 (92.3% of these valid TZs)
--- h Even = 79, h Odd = 64, h Unknown = 0
--- h = 30 [29], 31 [13], 32 [8], 33 [9], 34 [6], 35 [10], 36 [10], 37 [5], 38 [5], 39 [3], 40 [4], 41 [4], 42 [2], 43 [4], 45, 46, 47, 48 [2], 49, 51 [3], 53 [3], 54 [4], 55, 57 [2], 58 [2], 59, 60 [2], 61, 65, 72, 78, 80 [2], 137
--- where h is known and divisible by 6 = 51
--- hBase >= hTarget = 143
--- max retracement from TZ tip back to Base TZ tip (> 100% is beyond Base TZ):
-------- < 10% = 27 (18.9% of TP TZs)
-------- 10 - 19% = 55 (38.5% of TP TZs)
-------- 20 - 29% = 17 (11.9% of TP TZs)
-------- 30 - 39% = 20 (14.0% of TP TZs)
-------- 40 - 49% = 6 (4.2% of TP TZs)
-------- 50 - 59% = 2 (1.4% of TP TZs)
-------- 60 - 69% = 6 (4.2% of TP TZs)
-------- 70 - 79% = 6 (4.2% of TP TZs)
-------- 80 - 89% = 2 (1.4% of TP TZs)
-------- 90 - 99% = 2 (1.4% of TP TZs)
--- min retracement bars (min / max / avg) = 1 / 1 / 1.0

SL = 9 (5.8% of these valid TZs)
--- h Even = 6, h Odd = 3, h Unknown = 0
--- h = 30 [2], 32, 34, 39, 40, 41, 47, 54
--- where h is known and divisible by 6 = 3
--- hBase >= hTarget = 9
--- max retracement from TZ tip back to Base TZ tip (> 100% is beyond Base TZ):
-------- 100 - 109% = 1 (11.1% of SL TZs)
-------- 110 - 149% = 4 (44.4% of SL TZs)
-------- 150 - 199% = 1 (11.1% of SL TZs)
-------- >= 200% = 3 (33.3% of SL TZs)
--- min retracement bars (min / max / avg) = 1 / 1 / 1.0

EX = 3 (1.9% of these valid TZs)
--- h Even = 2, h Odd = 1, h Unknown = 0
--- h = 30, 33, 42
--- where h is known and divisible by 6 = 2
--- hBase >= hTarget = 3
--- max retracement from TZ tip back to Base TZ tip (> 100% is beyond Base TZ):
-------- 60 - 69% = 1 (33.3% of EX TZs)
-------- 70 - 79% = 1 (33.3% of EX TZs)
-------- 90 - 99% = 1 (33.3% of EX TZs)
--- min retracement bars (min / max / avg) = 1 / 1 / 1.0

Invalid TZs:
--- Base/Target TZ h Relation = 33
--- incomplete = 1

--------------------------------------------------
All TZs
--------------------------------------------------
Total = 1781
--- Up = 828 (46.5% of total)
--- Down = 953 (53.5% of total)
Valid = 919 (51.6% of these TZs)

TP = 839 (91.3% of these valid TZs)
--- h Even = 430, h Odd = 409, h Unknown = 0
--- h = 30 [29], 31 [17], 32 [15], 33 [16], 34 [13], 35 [18], 36 [23], 37 [17], 38 [13], 39 [10], 40 [25], 41 [15], 42 [12], 43 [15], 44 [7], 45 [8], 46 [6], 47 [13], 48 [8], 49 [6], 50 [7], 51 [7], 52 [8], 53 [15], 54 [13], 55 [8], 56 [6], 57 [8], 58 [13], 59 [14], 60 [9], 61 [8], 62 [7], 63 [5], 64 [5], 65 [9], 66 [6], 67 [7], 68 [6], 69 [3], 70 [5], 71 [8], 72 [6], 73 [4], 74 [8], 75 [5], 76 [3], 77 [9], 78 [6], 79 [3], 80 [5], 81 [3], 82 [3], 83 [6], 84 [5], 85 [5], 86 [2], 87 [4], 88, 89 [8], 90 [8], 91 [7], 92 [4], 93 [4], 94 [2], 95 [4], 96 [6], 97, 98 [7], 99 [3], 100 [5], 101 [6], 102 [6], 105 [2], 106 [2], 107 [5], 108 [2], 110, 111, 112 [3], 113, 114 [2], 115 [3], 116 [3], 117 [2], 118 [3], 119 [3], 120 [3], 121 [2], 124 [2], 125, 126 [5], 127 [2], 129 [3], 131 [3], 132, 133 [3], 134 [4], 135 [2], 136 [2], 137 [5], 139 [2], 140, 142, 143 [2], 144, 145 [2], 146, 148, 150 [2], 151, 152 [2], 153, 154 [2], 155 [2], 156 [2], 157 [3], 161, 162, 163, 164 [2], 165, 166 [2], 167 [2], 168, 170, 171 [2], 172 [2], 173 [2], 175, 176 [3], 177, 178, 179 [4], 180 [2], 181 [2], 182, 183, 184 [2], 185 [3], 186, 189 [2], 192, 193, 194 [2], 195 [2], 196 [3], 197, 198 [4], 199, 200 [3], 201, 202 [3], 203 [4], 204, 206, 208 [2], 209 [4], 210, 211 [2], 212, 213 [2], 214 [2], 215, 216, 222, 242, 243, 245, 248, 250, 254, 256, 257, 258, 260, 263, 268, 272, 275, 278, 284, 286, 288, 289, 293, 294, 304, 310, 315, 325, 326, 334, 336, 355, 356, 357, 358, 359, 408, 413, 422, 423, 437, 443, 444, 445, 448, 506
--- where h is known and divisible by 6 = 175
--- hBase >= hTarget = 839
--- max retracement from TZ tip back to Base TZ tip (> 100% is beyond Base TZ):
-------- < 10% = 288 (34.3% of TP TZs)
-------- 10 - 19% = 279 (33.3% of TP TZs)
-------- 20 - 29% = 109 (13.0% of TP TZs)
-------- 30 - 39% = 61 (7.3% of TP TZs)
-------- 40 - 49% = 33 (3.9% of TP TZs)
-------- 50 - 59% = 29 (3.5% of TP TZs)
-------- 60 - 69% = 16 (1.9% of TP TZs)
-------- 70 - 79% = 11 (1.3% of TP TZs)
-------- 80 - 89% = 10 (1.2% of TP TZs)
-------- 90 - 99% = 3 (0.4% of TP TZs)
--- min retracement bars (min / max / avg) = 1 / 1 / 1.0

SL = 60 (6.5% of these valid TZs)
--- h Even = 35, h Odd = 25, h Unknown = 0
--- h = 30 [2], 32, 33 [3], 34 [4], 35 [3], 36 [2], 39 [2], 40, 41 [2], 42, 46 [4], 47 [2], 50, 53, 54 [2], 55, 60, 62, 68, 69, 70, 76, 81, 82, 85, 94, 95, 96, 100, 102, 106, 109, 112, 118, 124, 150, 153, 174, 179, 185, 208, 223, 231, 409
--- where h is known and divisible by 6 = 12
--- hBase >= hTarget = 60
--- max retracement from TZ tip back to Base TZ tip (> 100% is beyond Base TZ):
-------- 100 - 109% = 3 (5.0% of SL TZs)
-------- 110 - 149% = 35 (58.3% of SL TZs)
-------- 150 - 199% = 9 (15.0% of SL TZs)
-------- >= 200% = 13 (21.7% of SL TZs)
--- min retracement bars (min / max / avg) = 1 / 1 / 1.0

EX = 20 (2.2% of these valid TZs)
--- h Even = 13, h Odd = 7, h Unknown = 0
--- h = 30, 33, 34, 41, 42, 46, 48, 63, 65, 71, 78, 80, 88 [3], 95, 96, 103, 116, 138
--- where h is known and divisible by 6 = 6
--- hBase >= hTarget = 20
--- max retracement from TZ tip back to Base TZ tip (> 100% is beyond Base TZ):
-------- 60 - 69% = 4 (20.0% of EX TZs)
-------- 70 - 79% = 4 (20.0% of EX TZs)
-------- 80 - 89% = 4 (20.0% of EX TZs)
-------- 90 - 99% = 8 (40.0% of EX TZs)
--- min retracement bars (min / max / avg) = 1 / 1 / 1.0

Invalid TZs:
--- before first opposite = 2
--- h lower than last = 336
--- Base/Target TZ h Relation = 521
--- insufficient retrace = 2
--- incomplete = 1

EXCELLENT WORK!!! Going over it a few more times presently!!!

Thanks Bill.

Before we get into the optimisation process, here's another report with realistic retracement settings. The last one had a minimum retracement of 0%, so all otherwise-valid TZs were included (except 2, or 0.2% of all valid ones, which I think is caused by a rare rounding error and isn't significant). That's why the TP rates in the last report are so high (> 90%).

Looking at that report we can see that 288 of the TPs (34% of them) retraced < 10%. Not good. But 279 (33%) retraced between 10 and 19%, and 109 (13%) retraced between 20 and 29%. After that the numbers drop off. So what happens if we specify a Min Retrace of 10%, and a Max Retrace of 30%? That's 2:1 R:R, or a required minimum TP rate of 67% to break even over time (assuming entry at 10%, SL at 30%, TP at 0%, no position management). Here's the report, with the TP rates highlihted in bold. As you can see, the actual TP rates are less than the required minimum. Not good!

=====================================================================
$EURUSD M240, Pip = 0.0001, Run time = 19-Apr-2015 17:24 (local)
h = 30-999+, h Variance = 0, k Min = 1 (applies to Target TZs only), Recheck = False
h Base/Target Relation = 'hBase >= hTarget' only, h Even/Odd = Both, h Div 6 = Included
Min Retrace (Entry) = 10%, Max Retrace (SL) = 30%, Min TP = 66.7%
Total bars = 9999 (from 12-Nov-2008 07:00 to 17-Apr-2015 10:00)

*** All h values here and on the chart are hMax (maximum number of left-transient bars in the TZ) ***

--------------------------------------------------
First TZs
--------------------------------------------------
Total = 189 (100.0% of all TZs)
Valid = 127 (67.2% of these TZs)

TP = 72 (56.7% of these valid TZs)
--- h Even = 39, h Odd = 33, h Unknown = 0
--- h = 30 [17], 31 [3], 32 [3], 33 [4], 34 [5], 35 [5], 36 [5], 37 [4], 38, 39 [2], 40 [2], 41 [2], 42, 43 [4], 45, 48, 49, 51, 53, 54 [2], 55, 57, 58, 60, 61, 65, 137
--- where h is known and divisible by 6 = 27
--- hBase >= hTarget = 72
--- max retracement from TZ tip back to Base TZ tip (> 100% is beyond Base TZ):
-------- 10 - 19% = 54 (75.0% of TP TZs)
-------- 20 - 29% = 17 (23.6% of TP TZs)
-------- 30 - 39% = 1 (1.4% of TP TZs)
--- min retracement bars (min / max / avg) = 1 / 1 / 1.0

SL = 55 (43.3% of these valid TZs)
--- h Even = 33, h Odd = 22, h Unknown = 0
--- h = 30 [12], 31 [5], 32 [2], 33 [5], 34, 35 [4], 36 [3], 37, 38 [2], 39, 40 [3], 41 [2], 42 [2], 47, 51 [2], 53, 54 [3], 60, 72, 78, 80 [2]
--- where h is known and divisible by 6 = 23
--- hBase >= hTarget = 55
--- max retracement from TZ tip back to Base TZ tip (> 100% is beyond Base TZ):
-------- 30 - 39% = 19 (34.5% of SL TZs)
-------- 40 - 49% = 6 (10.9% of SL TZs)
-------- 50 - 59% = 2 (3.6% of SL TZs)
-------- 60 - 69% = 7 (12.7% of SL TZs)
-------- 70 - 79% = 7 (12.7% of SL TZs)
-------- 80 - 89% = 2 (3.6% of SL TZs)
-------- 90 - 99% = 3 (5.5% of SL TZs)
-------- 100 - 109% = 1 (1.8% of SL TZs)
-------- 110 - 149% = 4 (7.3% of SL TZs)
-------- 150 - 199% = 1 (1.8% of SL TZs)
-------- >= 200% = 3 (5.5% of SL TZs)
--- min retracement bars (min / max / avg) = 1 / 2 / 1.0

EX = 0 (0.0% of these valid TZs)

Invalid TZs:
--- Base/Target TZ h Relation = 33
--- insufficient retrace = 28
--- incomplete = 1

--------------------------------------------------
All TZs
--------------------------------------------------
Total = 1781
--- Up = 828 (46.5% of total)
--- Down = 953 (53.5% of total)
Valid = 615 (34.5% of these TZs)

TP = 376 (61.1% of these valid TZs)
--- h Even = 179, h Odd = 197, h Unknown = 0
--- h = 30 [17], 31 [4], 32 [8], 33 [8], 34 [8], 35 [11], 36 [10], 37 [8], 38 [6], 39 [7], 40 [11], 41 [9], 42 [8], 43 [9], 44 [4], 45 [7], 46, 47 [5], 48 [5], 49 [4], 50, 51 [4], 52 [4], 53 [8], 54 [8], 55 [4], 56, 57 [5], 58 [5], 59 [10], 60 [4], 61 [5], 62 [3], 63 [2], 64 [2], 65 [5], 66, 67 [3], 68 [3], 69, 70 [3], 71 [5], 72 [3], 73, 74, 75, 76 [2], 77 [6], 78 [5], 79 [2], 80, 82 [2], 83 [4], 84 [5], 85 [2], 87 [2], 89 [6], 90 [3], 91 [3], 92 [2], 94, 95 [2], 96, 99, 100 [2], 101 [2], 102, 105 [2], 106, 107, 108 [2], 111, 112 [2], 115, 116 [2], 117 [2], 118 [3], 119, 126 [2], 129 [2], 131, 133, 135, 136, 137 [2], 140, 142, 144, 146, 148, 152, 155, 156, 157 [3], 162, 163, 166, 167, 171, 172 [2], 179 [2], 180, 181, 185, 189 [2], 193, 195, 196, 197, 200, 203 [2], 209, 214, 215, 242, 243, 250, 256, 263, 268, 272, 286, 289, 315, 326, 336, 355, 413, 448
--- where h is known and divisible by 6 = 80
--- hBase >= hTarget = 376
--- max retracement from TZ tip back to Base TZ tip (> 100% is beyond Base TZ):
-------- 10 - 19% = 263 (69.9% of TP TZs)
-------- 20 - 29% = 109 (29.0% of TP TZs)
-------- 30 - 39% = 4 (1.1% of TP TZs)
--- min retracement bars (min / max / avg) = 1 / 8 / 1.2

SL = 239 (38.9% of these valid TZs)
--- h Even = 139, h Odd = 100, h Unknown = 0
--- h = 30 [12], 31 [6], 32 [2], 33 [10], 34 [8], 35 [9], 36 [11], 37 [4], 38 [3], 39 [3], 40 [10], 41 [6], 42 [5], 43 [3], 44, 46 [7], 47 [5], 48 [2], 49, 50 [5], 51 [2], 52 [3], 53 [4], 54 [5], 55, 56, 58 [2], 59, 60 [4], 62 [2], 63 [3], 64 [2], 65 [3], 66 [3], 67, 68, 69, 70 [2], 71 [2], 72 [2], 73 [2], 74 [2], 75 [4], 76, 77, 78 [2], 80 [3], 81, 82, 83 [2], 85 [2], 88 [4], 90, 91, 94, 95 [2], 96 [3], 98, 99, 100 [2], 102 [2], 103, 106, 107 [2], 109, 110, 112, 116, 118, 119, 120, 121, 124, 127, 129, 134, 137 [2], 138, 150 [2], 153, 174, 179, 185, 186, 198, 200, 202 [2], 208, 210, 211, 213, 216, 223, 231, 258, 294, 359, 409, 506
--- where h is known and divisible by 6 = 63
--- hBase >= hTarget = 239
--- max retracement from TZ tip back to Base TZ tip (> 100% is beyond Base TZ):
-------- 30 - 39% = 57 (23.8% of SL TZs)
-------- 40 - 49% = 33 (13.8% of SL TZs)
-------- 50 - 59% = 29 (12.1% of SL TZs)
-------- 60 - 69% = 20 (8.4% of SL TZs)
-------- 70 - 79% = 15 (6.3% of SL TZs)
-------- 80 - 89% = 14 (5.9% of SL TZs)
-------- 90 - 99% = 11 (4.6% of SL TZs)
-------- 100 - 109% = 3 (1.3% of SL TZs)
-------- 110 - 149% = 35 (14.6% of SL TZs)
-------- 150 - 199% = 9 (3.8% of SL TZs)
-------- >= 200% = 13 (5.4% of SL TZs)
--- min retracement bars (min / max / avg) = 1 / 8 / 1.2

EX = 0 (0.0% of these valid TZs)

Invalid TZs:
--- before first opposite = 2
--- h lower than last = 336
--- Base/Target TZ h Relation = 521
--- insufficient retrace = 306
--- incomplete = 1

Finally, what about a wider retracement setting of 40%-80%? R:R is 1:1, need > 50% TP to be profitable. Nice and simple. Here's the report:

=====================================================================
$EURUSD M240, Pip = 0.0001, Run time = 19-Apr-2015 17:44 (local)
h = 30-999+, h Variance = 0, k Min = 1 (applies to Target TZs only), Recheck = False
h Base/Target Relation = 'hBase >= hTarget' only, h Even/Odd = Both, h Div 6 = Included
Min Retrace (Entry) = 40%, Max Retrace (SL) = 80%, Min TP = 50.0%
Total bars = 9999 (from 12-Nov-2008 07:00 to 17-Apr-2015 10:00)

*** All h values here and on the chart are hMax (maximum number of left-transient bars in the TZ) ***

--------------------------------------------------
First TZs
--------------------------------------------------
Total = 189 (100.0% of all TZs)
Valid = 36 (19.0% of these TZs)

TP = 21 (58.3% of these valid TZs)
--- h Even = 12, h Odd = 9, h Unknown = 0
--- h = 30 [4], 31 [2], 33 [2], 35 [2], 36 [2], 40, 41, 42, 51, 53, 54, 60, 80 [2]
--- where h is known and divisible by 6 = 9
--- hBase >= hTarget = 21
--- max retracement from TZ tip back to Base TZ tip (> 100% is beyond Base TZ):
-------- 40 - 49% = 6 (28.6% of TP TZs)
-------- 50 - 59% = 2 (9.5% of TP TZs)
-------- 60 - 69% = 6 (28.6% of TP TZs)
-------- 70 - 79% = 6 (28.6% of TP TZs)
-------- 80 - 89% = 1 (4.8% of TP TZs)
--- min retracement bars (min / max / avg) = 1 / 7 / 3.3

SL = 13 (36.1% of these valid TZs)
--- h Even = 7, h Odd = 6, h Unknown = 0
--- h = 30 [3], 32, 33 [2], 34, 35, 39, 40, 41, 47, 54
--- where h is known and divisible by 6 = 4
--- hBase >= hTarget = 13
--- max retracement from TZ tip back to Base TZ tip (> 100% is beyond Base TZ):
-------- 80 - 89% = 1 (7.7% of SL TZs)
-------- 90 - 99% = 3 (23.1% of SL TZs)
-------- 100 - 109% = 1 (7.7% of SL TZs)
-------- 110 - 149% = 4 (30.8% of SL TZs)
-------- 150 - 199% = 1 (7.7% of SL TZs)
-------- >= 200% = 3 (23.1% of SL TZs)
--- min retracement bars (min / max / avg) = 1 / 6 / 3.0

EX = 2 (5.6% of these valid TZs)
--- h Even = 2, h Odd = 0, h Unknown = 0
--- h = 30, 42
--- where h is known and divisible by 6 = 2
--- hBase >= hTarget = 2
--- max retracement from TZ tip back to Base TZ tip (> 100% is beyond Base TZ):
-------- 60 - 69% = 1 (50.0% of EX TZs)
-------- 70 - 79% = 1 (50.0% of EX TZs)
--- min retracement bars (min / max / avg) = 5 / 19 / 12.0

Invalid TZs:
--- Base/Target TZ h Relation = 33
--- insufficient retrace = 120

--------------------------------------------------
All TZs
--------------------------------------------------
Total = 1781
--- Up = 828 (46.5% of total)
--- Down = 953 (53.5% of total)
Valid = 180 (10.1% of these TZs)

TP = 88 (48.9% of these valid TZs)
--- h Even = 49, h Odd = 39, h Unknown = 0
--- h = 30 [4], 31 [2], 33 [4], 34, 35 [3], 36 [6], 39, 40 [3], 41 [2], 42 [2], 43 [2], 44, 46, 47, 49, 50 [4], 51, 52 [2], 53 [3], 54 [2], 58 [2], 60 [3], 62, 63, 64, 65 [2], 66 [3], 72, 73 [2], 75 [3], 77, 80 [2], 83, 85, 88, 91, 100, 119, 120, 127, 137 [2], 202 [2], 210, 211, 213, 216, 258, 294, 359, 506
--- where h is known and divisible by 6 = 26
--- hBase >= hTarget = 88
--- max retracement from TZ tip back to Base TZ tip (> 100% is beyond Base TZ):
-------- 40 - 49% = 31 (35.2% of TP TZs)
-------- 50 - 59% = 29 (33.0% of TP TZs)
-------- 60 - 69% = 16 (18.2% of TP TZs)
-------- 70 - 79% = 11 (12.5% of TP TZs)
-------- 80 - 89% = 1 (1.1% of TP TZs)
--- min retracement bars (min / max / avg) = 1 / 109 / 8.1

SL = 84 (46.7% of these valid TZs)
--- h Even = 48, h Odd = 36, h Unknown = 0
--- h = 30 [3], 32, 33 [5], 34 [5], 35 [4], 36 [3], 39 [2], 40 [3], 41 [3], 42, 43, 46 [5], 47 [2], 50, 53, 54 [2], 55, 56, 60, 62, 63, 65, 68, 69, 70, 71 [2], 75, 76, 78, 80, 81, 82, 85, 88 [3], 94, 95, 96, 100, 102, 106, 107, 109, 112, 118, 124, 138, 150, 153, 174, 179, 185, 208, 223, 231, 409
--- where h is known and divisible by 6 = 16
--- hBase >= hTarget = 84
--- max retracement from TZ tip back to Base TZ tip (> 100% is beyond Base TZ):
-------- 80 - 89% = 13 (15.5% of SL TZs)
-------- 90 - 99% = 11 (13.1% of SL TZs)
-------- 100 - 109% = 3 (3.6% of SL TZs)
-------- 110 - 149% = 35 (41.7% of SL TZs)
-------- 150 - 199% = 9 (10.7% of SL TZs)
-------- >= 200% = 13 (15.5% of SL TZs)
--- min retracement bars (min / max / avg) = 1 / 49 / 6.3

EX = 8 (4.4% of these valid TZs)
--- h Even = 6, h Odd = 2, h Unknown = 0
--- h = 30, 42, 46, 48, 95, 96, 103, 116
--- where h is known and divisible by 6 = 4
--- hBase >= hTarget = 8
--- max retracement from TZ tip back to Base TZ tip (> 100% is beyond Base TZ):
-------- 60 - 69% = 4 (50.0% of EX TZs)
-------- 70 - 79% = 4 (50.0% of EX TZs)
--- min retracement bars (min / max / avg) = 5 / 19 / 10.9

Invalid TZs:
--- before first opposite = 2
--- h lower than last = 336
--- Base/Target TZ h Relation = 521
--- insufficient retrace = 742

As you can see, 40-80% is looking better, but it's nowhere near phenonomal. It was time to start optimising those parameters I mentioned above.

More next week. Have a good weekend (what's left of it)

Thank you, NorthTrader! Excellent work indeed!

Thanks SputnikX. I'll be away from the thread for a few days because I think I've spotted a flaw in the logic of one of the parameters. It concerns "h Base >= h Target" stats. If so, it will change the probabilities posted above. For the better or the worse? I don't know. Back later.

It is really refreshing to see someone doing some work here and actually producing some stats to back it up. Thank you for sharing.

I disagree, lemna! My work is just a drop in the ocean of some of the fine analysis that goes on here at FF, whether it's done automatically by programs or manually by inspecting charts. But thanks anyway

Problem update: the code was written exactly as we specified it, based on lengthy discussions following chart analysis and previous results. However, it seems that not all of the first opposite TZs are being included. There is a discrepancy between the double dequence stats posted earlier in this thread - which are a collection of fixed-h runs - and the variable-h reports posted later. This discrepancy arises when the Max h of the Target TZ is greater than the Max h of the Base TZ. Instead of excluding these setups, they should be included with the proviso that the actual h used (to specify the time limit) is no greater than the Max h of the Base TZ. This would then match the logic of the double sequence stats.

Of course, there are many ways to use TZs and the way it is coded now may turn out to be the better one. So I won't alter the current functionality. What I'll do is add a new option to report them exactly as the double sequence indicator does, but for a range of h values. This may or may not produce better results. Let's wait and see!

Hey NorthTrader, Hope all is well. Still following the thread with great interest. I am interested to see how your change in the code will change the results of this. I do not want to sound like a sour puss but I will not be surprised if the more you increase your sample size the closer to 50% the hit rating will be. It will like favor bull or bear setups if price is closing above/below where the sample size started also btw, this is something I have noticed in my experience testing this stuff, and this is one of the only few things that really provides any setup with an edge over another one. I have done extensive statically modeling on a number of different methods and every single one boils down to 50/50(minus spread or if price sample is ending above or below where the sample size started, agin if this is the case bull/bear setups will be slightly favored over its counter part) in the long run. It's just the nature of the market. Just like there is a 100% chance all levels price has been, will be returned to eventually. Every level will be recurrent if you make the h long enough.

I still think the method you are using can be very valuable trading but I do not think the real edge will be in the setup you are chasing, it will be in how the setup is traded, which is the case for any signal or entry method. This is why some can trade a system successfully and others cant. Can't wait to see what else you come up with as I love threads that include real hard data and idea's, not just a bunch of mumbo jumbo out of reach stuff. these are just some of my thoughts on the process and what you have posted so far.

EDIT:

Here is an example of what I mean. Below is the RSI(14) readings distribution for EURUSD since last September for the hourly chart. Since price has decreased over the sample period the oversold levels are obvious going to be favored as any bear signals will be.



One thing to notice is the 14 period RSI readings distribution forms is the classic bell curve. This is how it looks for all timeframes. There is a method to the madness of the default 14 period.

One more thing I wanted to share in regards to how i believe the probabilities will work out. The current range you are getting of 40-80% that I eventually think will even out to 50/50 is because of Bernoulli Distributions and size of your sample. If i read the stats right based on the tradable retraces there were only 36 valid first TZ's and 180 Valid All TZ Setups which is why I believe there to be a huge range in the win rate.

"In a Bernoulli trial, the outcome is random and can have one of two values: "yes" or "no", "boy" or "girl", "heads" or "tails", and so on. The variance of a Bernoulli distribution is: Var(X) = np(1-p) where n is the number of trials (100 in this case), p is the expected probability of outcome 1 and (1 - p) is the expected probability of outcome 2. Since we expect the probability of a coin landing on heads to be 0.5, the variance of a coin toss experiment with 100 trials should be (100)(0.5)(0.5) = 25. The square root of 25 is 5, so if you repeated a 100-trial coin toss experiment over and over, the majority of them (about 68%, assuming a fair coin) should yield between 45 and 55 heads (45 to 55%). Only about 32% of the experiments should produce fewer than 45 or more than 55 heads, and only about 4% would be expected to produce fewer than 40 or more than 60 heads. Just to reiterate and crunch what I've already said, if you flip a fair coin 100 times, there's a 68% chance that the number of heads will be between 45 and 55. Any variation within that range should be considered COMPLETELY NORMAL!

Now let's say you flipped the coin 10,000 times. That experiment has a variance of (10,000)(0.5)(0.5), or 2500, and a standard deviation of 50. Although the standard deviation is larger it was in the 100-coin experiment, the sample size is MUCH MUCH larger. The majority of your 10,000-coin toss experiments should produce between 4950 and 5050 heads, or between 49.5 and 50.5%. That's much closer to giving you the expected 50/50 distribution than what you'd get with a 100-coin toss. If you increase the number of trials to 1,000,000, you get even closer to the 50/50 distribution: most million-coin experiments should fall from 49.95% and 50.05% heads."

With a Samples sizes of 100 flips you can see that basically 1st std deviation is 45 and 55(5% increases) and the 2nd std deviation is 60 and 40(10% increases). This is for normal distributions which the market does not do. I can post a study done showing that through out history the standard deviations will be up to 300% or larger than what is "normal". and again this is because markets are far from normal.

In other words if this coin toss in the example were the market, the 1st standard deviation could really be, remember this is at extreme peaks and valleys of the market, 35 and 65. and the 2nd std deviation would be around 80% and 20%.

Hope this stuff helps in your work! Again please let me know if I am posting too much in here or if these post seem off topic. Some interesting stuff to consider if you havent looked at Bernoulli trials already.

Yes, you're right about small sample sizes. We have to be very careful about them. M240 is a double-edged sword - nice setups, but difficult to analyse accurately due to the low number of them.

You're also right about most of the edge being in how you trade the setups, not in the setups themselves. It would be naive to think that you could always enter at 40% retracement, SL at 80%, and hit TP more than 50% consistently. If you could do that then you would simply run it automatically 24/7 and become very rich. It's not going to happen! That's why I think dynamic retracement levels are the way to go. The market is dynamic, so the levels must also be dynamic to match the current conditions. And another idea is to trade lower timeframes within the context of the M240 setup.

So what's the point of all this fixed-level analysis? I suppose initially it's to prove how feasible, or unfeasible, the system basically is. Anything approaching Min TP rate for a range of R:Rs isn't bad in my opinion. Much less than Min TP rate would be bad. This also builds confidence in trading the system, because if price isn't too far from the TZ, and the trend is in favour of the TZ, and market conditions seem "normal" (discrection and experience required), then you know from the stats that there's a good chance of the TZ being resolved.

And once the foundation has been coded, you can add and optimise any number of parameters to see how they alter the results. You can also see the setups and levels marked on the chart and further analyse them that way. There are a few which you wouldn't take because they don't look "right" according to the market environment at the time. The indicator doesn't know that, but you do.

Problem update: the flaw wasn't as bad as I first thought. With the h Relation parameter set to "All" (different to the above stats), it almost emulates the Double Sequence stats perfectly. The only thing it doesn't do is use the h value of the Base TZ for the Target TZ if Target h is greater than Base h. It must do this to match the logic of the Double Sequence stats.
It also stops checking after the first TZ is found that has an h value greater than the h Max parameter. Why? Because most of us probably don't want to hang around in a trade longer than a few weeks (I don't, anyway). So what's the point of checking more than, say, h=120? (4 weeks on M240)
Having said that, I'm going to add another parameter to continue checking Target TZs if all the previous ones were invalid due to Min Retrace, even if their h is greater than h Max. This goes against the logic of Double Sequence, BUT it will produce a larger sample size and maybe interesting results.
I've also updated the chart markings to show the Base TZs, and their h values, more clearly. Will post an image later.

More stats this weekend hopefully Bye for now.