Channel-to-pivot approach

Fellow traders,

As a tribute to the many ideas and insights I gathered from this forum, I'd like to share an approach that is serving me well. Ahead of explaining the approach, let me be clear: even though I have developed/coded all necessary tools by myself and they are readily available, I do not intend to share (nor support) them, so don't bother asking me for them. I'd be going into all their relevant inner workings, so if there is anybody interested in pursuing the approach, it should be easy to reproduce them.

Trading Context
I am using a combination of regression calculations on currency pairs and related indices to get an overall directional bias, where the pair is heading (=trend). I am well aware of the non-stationary character of time series and the questionable value of statistics derived from them. Nevertheless, it is always a perception we are trading, and as such those regression channels help me build mine.

I built an indicator for each of the 8 currency indices from bar zero to 1440 on a 60 min chart (=3 months). The formula for the usd index is well-known: usd= ((USDCAD*USDCHF*USDJPY)/ (AUDUSD*GBPUSD*NZDUSD*EURUSD))^1/8. For the aud index it is aud=usd*AUDUSD, for cad=usd/USDCAD, etc.

It further derives the best fit by iteration. This is done by running linear regressions and tracking the coefficient of determination R^2 over all the 1440 samples per index starting with a minimum of 72 (=3 days). The algo will stick with the sample size with the highest R^2. The consequence is that each currency index ends up with an individually optimized sample size rather than applying the same "look-back period" across all of them.

The third and last step is then to run the same iterations of regressions across the traded currency pairs and compare the resulting biases with the ones of the relevant indices - e.g. CHFJPY bias with CHF and JPY biases. If they all align, e.g. JPY up (=strong), CHF down (=weak) and CHFJPY down, I would only looking to place sell orders.

see attachments

Entries
For entries (and exits) I pick a straight-forward pivot indicator showing Pivot, S1-2 and R1-2 levels. You can place pending orders on those levels above (in case of a short bias) or below (in case of a long bias) current price, but they should be either close to the line-of-best-fit (LBF) or better in pull-back zones - i.e. above LBF for shorts and vice versa for longs. In case the entry levels are not evenly spaced out, I might manually follow previous S/R to achieve that. I prefer even distributions to avoid putting too much weight on one level within a channel.

Exits
Similar to the entries, I will pick the pivot levels for exits, as well. The relevant targets should be at/below the LBF for shorts and vice versa for longs depending on your guts to hold the trades. Once the overall position (=swarm of orders) is in profit you can place a breakeven stop loss, if you like. I usually do this, when I have to step away from the computer. I don't trail individual orders and aim to close swarms during one trading session to benefit from higher balances for the next opportunity.

Risk management
The regressions and their standard errors give me a reasonable gauge for lot sizing - i.e. a std error of 15 pips, a capital of 10,000 and a setting of 0.5% would result in 0.33 lots or 3.30 dollars per pip on my account. I don't use stops but average-in with the same lot size, because I am trading a channel bias with some temporary statistical relevance. There is no singular truth about direction, but the odd relevance of pivot levels, the average price of an order swarm and continuous risk exposure. Assuming the unlikely situation of a 6 sigma move (=the huge parallel shift in the size of a whole channel) from the first entry with 5 more re-entries, you'd be facing an overall negative float of 0.5%*(6*7)/2=10.5%. Since I track every order individually, that's when I'd close that particular order with a loss of 3%. Unless there is a Brexit or SNB unpeg type of event with lasting tectonic shifts, there is usually enough volatility to unwind the swarms within a few days at max. I'd like to stress here, that I'm deliberately trading with more risk than others, because I'm in it to generate continuous income.

I'd be glad to answer your questions.

Please share your indicator

One setup this morning on NZDUSD.

CAD has been steadily strengthening over the last three days (about 74 hours) and it eventually changed bias altogether from 13 hours ago. Even though I had been observing the slope change since about 2 days, I still decided to enter 2 trades in the old direction yesterday. Have a look at the attachment. There was still ample volatility to get out, as long as you play the levels rather than wait for a huge continuation trade.

MathTrader7 has created a nice tool for offline index charts. It is a lot more than you need.
As for the regression channels, you can start by playing with the MT4 built-in regression tool.
Pivot point indicators are all over this forum, just run a search.

Interesting, I used a Regression technique when trading binary options a few years ago and had a fair bit of success with it, i will keep an eye on this and see how it goes.

Good Luck to you!

Thanks.

The key is to avoid being on the wrong side when the bias is changing/ranging.

1. You should gauge the pull-backs / adverse excursions from the LBF in the context of currency pairs and individual currencies: I wouldn't pick a pivot level for first entry that is outside the channel and/or adverse excursions of related indices are extreme (e.g. |z-score|>2.5).
2. The day-to-day pivot levels must show continuation of "staircases". Assuming the bias is down, the current R2 would have to be lower than the previous one and vice versa for S2 during upward bias.
3. You can also run 2 additional regressions over the pair of interest, but this time with the optimal sample sizes derived from the individual indices. As a result you can see the appearance of adverse drifts on the shorter term or a flattening of slopes of the LBF. That would make me cautious, too.

OMG, I see myself in the mirror in some of your thoughts showed here! I have never thought I'll say that to anyone.
Is the mix of pivots and regression and deviations - parallel channels (I use Fibo channels Tool to extend the deviations)
Well, regression, so unfortunately underrated as a bias detector tool! ...fools!
I guess I'll have a test of your sys. for better profit ratio, better than my current systems.
Thank you, keep it up!

(I'll silently follow this thread)

Good to see like-minded people, Tudor.

Pivot levels are probably more related to median lines than to regressions, which still doesn't explain their oddly predictive value in the world of Forex (vis-a-vis non-24 hour markets). I have been struggling to find the optimal sample size for a long time now and at some point thought inflection points to be the solution (hence my avatar..). However, in a cosmos of unpredictable swings, it is not the true signal that matters but your perception and method how to deal with it. Only then I started looking at regressions and data similarity/dispersion in combination with the only somewhat predictable levels I could find - i.e. pivot levels, rounded numbers, etc.

Today I entered EURGBP short - 3 times. The last one still running.

I use to trade deviations at the (linear) regression direction and angle.
The main deviation was parallel to the regression line and tangent to the top/low wicks or Close points (H1). Used that to mark TP/SL.
Sometimes I have extended the deviations drawing more parallel on the regression line (using Fibo channels tool)
and have my charts to look like yours, with multiple parallel chanels.
Here's a sample of the simple version, no multiple channels:

I just noticed gurangax' trade stats - how incredibly lucky he has been. He must have a whole army of angels watching over him. Those kind of results give us survivorship bias - a bit like relying on regressions..;-) There is zero risk management there, just pure hope for a beneficial future - and there was for now! That made me think over my own good stats: my risk management needs further refinement. I'll probably implement 7Bits snowball approach but not following a rigid grid structure - i.e. I'll have the EA manage adverse closures while I (pre-)define the entry levels. Time for some more coding.

Happy Easter.

From a directional momentum point of view, it is interesting to observe the confluence of the 38.2% fibo retracement level (in down trends) or the 61.8% fibo retracement level (in up trends) with the channel limits (I marked both in blue in the screenshot). The closer the distance between the mentioned fibo levels and the maximum excursion levels (i.e. 3 standard errors from LBF), the stronger the bias it seems.

P.S. the sloped lines do not represent the LBF but the channel limits - i.e. = LBF+3xStdErrors for downward bias in this case of EURGBP

...and with the pivot levels

...and going with the practice of self-fulfilling prophecies, I'd be picking entry zones where there is confluence between Fibonacci retracements and Pivot Points.

Target exits are the next x levels down.

Stop levels are either a few pips above the entry levels (for the sniper traders) or above the first Fibo retracement level above the channel limit for channel traders.

Despite the EURGBP chart shown above, my tools actually suggest NOT looking for market entries today: JPY is the only strong currency and all JPY pairs are oversold. I'll place some pending orders, though, and call it a day.

First entry today

It's easy to determine the bias strength. And the entry levels. No need for sophisticated techniques.
But (my) problem was always to find a proper hedging sequence to compensate the fakes, so many fakes

Indeed, there are no guarantees that the bias continues, but that's where risk management in the context of a channel comes in. My lot size is a function of the balance and the channel width (the distance between first entry and channel border are often >200 pips), and, e.g. in the case of a downtrend, the stop-loss of the lowest/worst order would end up being above the channel. Each order has its own stop-loss at x% of the balance but all of them share the same target. I enter new orders as long as the pair bias remains. Remember, there are 3 biases: each currency index and the pair itself. I don't manually close orders unless the pair bias has changed or I'm in crisis mode with little tolerance. In the worst of cases all my orders get stopped out. At up to 5 orders that can be 10% of my account, but those orders can as likely turn into a very powerful profitable swarm of 5-10%, because bias doesn't change that fast.

Personally, I wouldn't hedge, unless it is a natural one through another currency pair - e.g. I'm currently short GBPJPY and indeed the index bias for GBP has turned very strong within a few short hours; I could consider picking a GBP pair with a clear upwards bias such as GBPUSD.

Yes, is the classic risk management. Whatever your Risk Reward is, and doesnt matter how many trades you have opened,
the total amount of projected losses (SLs) should go around the maximum risk you allow yourself.