Can a great programmer make an EA from this?

This is a sample of a pdf I found on another post here at FF:

For example, when USDCHF moves 15 pips up or down on the day interval, the probability that it will continue to +40 (+25 net) is 89.9%.

Could an EA that monitors one pair only, be programmed to look for a move of 15 pips up or down on the 1D TF and then place an order going in the same direction with a TP of 20 (leave 5 for spread + slippage)?

Could you make it so that the pair can be changed, the TP can be altered and the movement in pips to trigger the trade can be altered to test its effect on other pairs analysed in the pdf?

Thanks for any help in advance.


Have attached the pdf,apologies but i'm not sure which thread I found it on.

I tried checking a few tutorials on programming an EA myself, but it went over my head like life goes over Gary Colemans (what you talkin' about Willis?)

I've been looking for this.

This is more or less the basis of renko trading.

The problem is, we don't know the time it took to move x pips in either direction. We don't know if the 10 pips in either direction were continuous, or choppy. If this truly were a price-only analysis, then why are there different percentages for each timeframe? I suspect there's something about this report the authors aren't telling us.

is there some thing wrong with the PDF attached?


my browser is getting closed when i try to download attachment

Please immediately forget this PDF.

The only explanation how the author could have arrived at these ridiculous numbers is that it all is based on the assumption that you never ever use a stop loss. A quick look at any chart will show you how the reality looks like.

The very way these meaningless numbers are presented (their sheer meaninglessness alone) and how the whole experiment is described shows that the author has never ever had any contact with statistics, he does not even seem to know somebody who's cousin knows somebody who ever had anything to do with statistics or any other form of serious research at all.

I'm not totally sure whether this document is only a testimonial of gigantic ignorance, or if it is meant as some kind of tasteless april fools joke or if there may be even fraudulent intentions (selling of an EA) involved.

Actually if you'd read the paper, the author did say they used a 20 pip stoploss. Nonetheless, verifying (or debunking) such a high hit rate is easy.

A solution *does* exist to the premise "if price moves x pips from point z, it will move another y pips without returning to point z n% of the time"

All you have to do is perform a statistical study to come up with answers that have a positive expectancy.

Ronald Raygun, I have read loads of your posts and seen how quickly you can make quality EA's.

If you have the time, could you maybe spare a little and knock up an EA to test this theory.

I would run it and post updates for my results on a weekly basis.

You could have daily updates as you are the one doing the hard part lol.

Fingers crossed you get time, but I imagine you get more requsts for EA writing than Jordan gets requsts to shut her mouth!

Still can't get my head around the fact you are only 19!


cheers Forexell (jason)

Thought I would point out that I found this pdf on this site, I don't have any connection with the author.

I noticed a comment regarding someone trying to maybe sell an EA, if they are, that guy aint me.

happy pipping

Then my first sentence must read: "There is no rational explanation of how the author could have arrived at these ridiculous numbers" instead of "The only explanation [...]"

I have read the paper a few months ago and i did what i recommend to others: I just forgot it.

This is mostly why I agree with 7-bit


Quote
--------------------------------------------------

To create this brief, we employed the following assumptions:
• The data is tic data from January 2000 through May 2004. Day intervals include data through October 2004.
• Each interval includes at least 5,000 observations except the day interval. In that case, we included each daily closing price from January 2000 through October 2004.
• We studied the following intervals, in minutes: 5, 30, 60, 120, 240, 1440.
• We studied the four major currency pairs: EURUSD, USDJPY, GBPUSD, USDCHF.
• We used a “trigger value” to set the entry condition for each simulated trade. To do this, we took each interval close and observed the next 24 intervals. If the trigger was hit in either direction, we entered the market. From that point we measured the probability of hitting our continuation price targets within the 24 periods.
• We employed a stop value of 20 pips for every simulated trade.
• No spread is taken into account.


---------------------------------------------------
End Quote



I highly doubt that this report is done on a tick-by-tick basis. Based on what the part in bold says, they checked the bar of the trigger price, and the next 24 bars after that to see which TPs or SLs were hit.

The part in red is a real game-ender because if you add in spread, your net gain is really 4-5 pips, not 10 (at the first level). If you really think you can prove me and 7-bit wrong, get a straddle script/EA (they're all over the place) and start testing.

This is data from 2004. Market dynamics have changed a lot since then. 2004 was more or less part of the heyday during which sub-prime trading was supreme and considered the new 'in' thing to do. What some funds started doing was borrowing money in Japan (where interest rates were low) and buying these securities in the US (where return on investment is higher than the interest rate). The difference between interest is what these firms kept, and--while the market was good--they were living well.

Best of luck!

I agree on the spread issue RR. The way the paper is written, it leads the reader to think it can be traded directly. But it's not enough to trade alone.

I still contend that one can find statistical edges this way and combine them with other statistical edges to form a profitable system.

For instance, instead of "straddling" the market, one could trade with a long/short bias based on another edge. A simple 20 EMA could be used. If price is above a 20 EMA, then the close will be higher than the open 60% of the time for 4 hour bars.

If you could find a couple of orthogonal edges and combine them, you can create a profitable system.

Do you have more such examples (preferrably othogonal ones)?

You guys have been doing this for some time now, i'm fresh to market.

Experienced traders like you tell me something is a lost cause, I listen.

Thanks for taking time to read what I asked and for all your considered responses

Sure. Candlestick patterns can be easily analyzed statistically. Check this site out. I wouldn't take this guy's percentages at face value but would do my own analysis in the market I was interested in. Still it's very compelling.

Something as simple as open-close patterns can be found which have a statistical edge.

I haven't looked at this yet, but supposedly pivot points do show their "fat tails" as well.

A fellow I know is looking at fib retracements. Think about this: if price moves n pips and retraces 38%, what's the probability it will reverse and continue? what's the probability it will reverse and continue after a 50% retrace?

The world is full of statistical edges. A creative person can find all sorts of places to look for these.

When I said orthogonal before, I meant that you'd combine edges from observed effects that aren't coupled. For instance, it doesn't make sense to try to combine edges from MACD and a moving average cross. Why? They're both based on moving averages so you're not really combining edges, just averaging them together.

On the other hand, if you determined that certain times of day or certain days of the week produce larger directional movement, combining this with a moving average and candlestick pattern makes sense.