Your Trading Edge Dilema?

A conversation I had with another trader yesterday reminded me of a story from which there are interesting questions that arise. I'll briefly recount the story before posing the questions.

I knew of a trader who thought he had found the goose that laid the golden egg. He had devised a trading edge (essentially a set and foreget B/O system) that when back tested performed well over what he reasonably thought was an extensive and representative sample. Yes there were losers but the expectancy was positive and there seemed every reason to go live with it and to leverage up aggressively. And so he did and for some time all went well. His aggressive use of leverage and compounding of retained gains made for an impressive account gain. He was not just leveraging proportionate to the retained gains but increasing leverage (position sizes) as the live results seemed to bear out the back test. Then distsater struck, a run of consecutive losses so severe that it resulted in him abandoning the trading edge. This run of losses was bad enough on it's own but was made worse by the increasingly larger position sizes such that at more sensible position sizes commensurate with the historical strike rate, the resulting drawdown on a/c might have been easier to live with but because of the aformenetioned it was, from memory, even aftyer redusing position sizes following the first detsabilising drawadown in excess of -60% at which point live trading stopped. His monitoring of the edge therefater showed that it would have crashed and burned totally and fairly quickly.

The original money management was I imagine based on the back test but I have no idea whether this took into account the most likely size of a long consecutive losing run- should it occur or indeed the probability of it occurring as oppossed to what had occurred. This is an important distinction It seemed to me that whilst the possibility was entertained the backtest had shown that it was not a high probability.

As the origantor acknowledged, there was one of the mistakes, the historical evidence showed that such a detstabilising consecutive losing run was a possibility but had not occurred such for it to be considered improbable in the future. A sample of data in the past can only point to what happened in the past but the possibility of a destabilising event remains howsoever improbable based only on historical data.

Clearly the other main mistake was the over leverage such that when the improbable (based on historical data) did occur it caused losses that detsabilised him and his ability to carry on. nIe his management of h=the trading edge exposed him to an unsees set of circumstances that he had no prior knowledge of how he would react to..In the evnt they were sufficiently detsabilising to warrent his abandonment of the Trading Edge.

Ther third mistake was that effectively the same bet was placed across correlated instruments so that instead of hedging the risk, the same bet was being placed that assisted in magnifying the losses.

So the questions are, and there may be more;

a. Do you know the probability of a consecutive losing run of trades at your strike rate, over and above that that has been experienced in a back or forward (live trading) sample ?

b. Is your money management such that if the worst should happen it doesn't expose you to any destabilising losses ?

c. Do you increase your position sizes in winning runs or decrease them in losing runs ? or any other methodology in this respect ?

Discuss.

a) This is not something you CAN really know. You can calculate the probability of winning and losing based on past results. But that isn't the ACTUAL probability of winning. You never know the real probability, and you can't know, because you don't even know if it is constant, the factors that determine the probability or even if those factors are constant. Moreover, you'd expect them to change over time.

So sure, if you calculate the historical probability of losing is 40%, then you can use a formula to try to calculate the probability of 15 losses in a row, or 14 losses out of 15 etc even though in backtests you might not have had more than 7. But that formula will likely assume your results are independent, which they aren't, and as mentioned they're not even on the real probability. So of how much use are they?

b) To say that you aren't destabilised would mean that you have to decide there can't be any more than X losses in a row and adjust your risk accordingly. But it is possible to have more losses than whatever you pick as X, so how can anyone say that they can't be destabilised by losses, unless they're trading demo? Also you can be destabilised in less trades than it takes for your account to be severely damaged. After 3-4 losses in a row, then you take the next trade and it's in profit. You should hold on and go with the system, but there is going to be a temptation to snatch some of those profitsoff the table, just because of only 3-4 losses. So you have to know how much you personally can handle.

c) Risking a fixed % means that you are trading larger size after winning, and smaller size after losing. But if you mean you normally risk 1% and then you go on a good run and start riskign 3%, 5% etc, then no. You could very quickly wipe out your good run with that.

Just my 2 cents

I'm working on a shorter term system which is showing some reasonable results BUT not quite so impressive once you strip out slippage and commission. The idea is to run it alongside my other system, as long as testing shows that risk-adjusted returns for the two systems are better than for either the individuals.

I've been down this route before .. my "second" system turned out to be too correlated to the main system; whilst it net made money, the rise and fall in equity was too similar hence I've had a re-think.

With regards your trader buddy, you've always got to start small, just to see how things go. Typically, when you encounter a good run of trades, that's the time to REDUCE stake size, not increase - I would think very carefully about increasing bet size in % of equity.

By inference, a good time to increase bet size might be when you're in a hole - but this may be the idiot's path. Why not just keep bet size constant, either in £ terms, or in % terms.

This is a bit of a ramble ---- but I think my point is that there probably isn't a singular methodology that will keep you happy, so diversification comes not in adding markets (which I think is what the protagonist probably did) but in adding systems.

Yes, I am going to end up like ODT, running 48 algos, having spent millions to develop them

bbmac said:

a. Do you know the probability of a consecutive losing run of trades at your strike rate, over and above that that has been experienced in a back or forward (live trading) sample ?

Click to expand...

It is possible to calculate the probability for any number of consecutive losing trade sequences based on the performance of any set of results generated by any trading method. And most stat-heads will caution you that any attempt to press probability into service on a trading method which hasn’t had sufficient time to develop a reasonably static performance profile is going to be pointless. The stats will simply not have any basis for consideration.

I’d go a little further than that and say for even any trading method which has had sufficient time to develop a reasonably static performance profile, it is going to be equally pointless.

Knowing that for any given profile the probability of getting a run of 3, 10 or 15 consecutive losses will not in any way convince reality to conform to those statistics. And even if that were not the case, I don’t believe there is any performance profile that delivers at such a regular rate over any statistically valid period of time for which statistics can be sensibly applied anyway.

Your W:L of 80:20 applies to how many of your last X trades? What’s your W:L over your last 1000 trades, last 100, last 10? It’s not the same, is it?

So trying to plan for specific actions to take if you hit more consecutive losers than you ‘should’ have had is folly. I could get my next 10 consecutive trades going into the red on a system for which based over the last few thousand trades, it really shouldn’t happen. But if it did, would I stop and consider the system to be suddenly flawed? No. Reality clusters well outside the scope of basic probability.

bbmac said:

b. Is your money management such that if the worst should happen it doesn't expose you to any detsabilising losses ?

Click to expand...

I risk 2% of my total trading capital on any given trade. On a longer run, I’ll potentially go in more than once on the same instrument. I won’t have more than 10% risked in the market in total at any given time. Losing 10% wouldn’t destabilise me. I’ve lost 100% on more than one occasion which is why I have the above rules now in place.

bbmac said:

c. Do you increase your position sizes in winning runs or decrease them in losing runs ? or any other methodology in this respect ?

Click to expand...

My position sizes are fixed at 2% regardless. If I get an entry setup on an instrument for which I already have a position (and if I’m not maxed) I’ll go in again, because it is appropriate to so based on market structure and my trading method indicates I could, not because ‘I had a good week last week’.

If I’ve been hitting good trades, my increasing capital will automatically increase my position size, as it will automatically reduce my size after a few bad ‘uns. A fixed percentage takes care of this for you.

As for winning runs and losing runs they don’t feature at all in my trading strategies. The number of consecutive winners or losers has no independent significance. The only impact they have upon my trading is in the amount by which they increase or decrease my total trading capital and the resultant size my 2% risk assumes on the next trade.

In the first "Wizards" book, a bet size of 0.5-1.0% was mentioned by several of the traders, it was perhaps one of the few common denominators. The styles were all wildly different.