Strategy Development: Modelling

[Cards37]

I was originally scalping and large moves killed any profit I made. So I then started to look at why the large moves occurred.

Pretty much exactly where I am now, its actually understanding the latter that is the issue!

[Derek27]

I've never bothered looking at why these big moves occur. I just accepted they occur and avoided all but the most stable markets.

[ShaunWhite]

I've never bothered looking at why these big moves occur. I just accepted they occur and avoided all but the most stable markets.

I see Mr Wing-It is hanging around in the strategy development thread No disrespect but I didn't think this topic was your scene. No need to reply to that, I'm trying not to push my question to Euler too far back.

[Euler]

Pretty much exactly where I am now, its actually understanding the latter that is the issue!

There is a balance to be had. I figured right at the very start of my career I could pick off opportunities, it was the losses that were a problem.

I then started to flip and flop strategies depending on what I thought should work. You can cut and refine entries to the point where you don't make any. So it's always a balance of risk and reward.

You don't have to trade every market for sure and you can pick odd the odd one or two, but finding a way of profiting from most is how you take trading to the next level.

[xitian]

You don't have to trade every market for sure and you can pick odd the odd one or two, but finding a way of profiting from most is how you take trading to the next level.

I think that’s a pretty important point. I’ve found the best strategies have the simplest principles behind them. As soon as you start tweaking, fiddling, filtering, you’re more likely to end up with something that either places very few bets or just doesn’t work at all because you’ve overfitted to past data.

Have a good reason why something should have a small edge (above random trading) that’s applicable in as many runners or situations as possible. Backtest the idea, and if it doesn’t work, move on to the next idea.

I recently found an edge that’s so simple I could describe the entire strategy (with parameters) in a sentence. That’s not to say it was easy to find. I might go months and months without finding anything new and profitable, but often when I do it’s quite simple.

[ShaunWhite]

assumes that almost any strategy fits a certain set of markets that can be singled out in a logical way with the data we have at our disposal.

A trader could keep on adding & adding variables & eventually, something will look amazing - but still provides little real predictive value - & that's my issue

I agree 100m% (a phase I dislike but sums up my feeling). And worse still when no amount of variables singles out the best results.

And re the variables, we also need to remember that Peter says the reason why a strategy is effective needs to be understood, rather that it just applying to an abritary subset. Find the variables (if indeed they exist), then find out why they worked.

I find it almost impossible to believe the adage that every strategy has it's market. If people are starting with pseudo random starting strategies and finding definable scenarios that work well enough to then refine still further, I think they've either been remarkably lucky or I'm still missing something.

I'm at the stage where I'd give my data & my strategy to a Peter or a Dallas or a Linus or a Jimi just to see what they found in it, or not. Even giving away something potentially valuable would be worthwhile to learn the key to unlocking more.

An actual worked example of a strategy that produced marginal results (ie one that isn't a trade secret and probably not worthh pursuing) would be great.

'Every strategy has it's market', really? I hope we can take this topic further when it's not quite so busy.

[ruthlessimon]

I'm at the stage where I'd give my data & my strategy to a Peter or a Dallas

I hope we can take this topic further when it's not quite so busy.

I think you'd have to be a pretty special trader to find something Peter & Dallas weren't already aware of. I'm actually in the same boat - I've found what I think is a ground-breaking result on certain markets; but the ramifications of the discovery means that I cannot trade certain (what many people would call) "bread-&-butter" markets. My quest to solve how to capture the "easy" markets, while maintaining the edge - keep failing again & again & again

That's why I'm biased against execution psychology Shaun - I'm looking for help just as the next man; & love talking about the hard quantitative aspects of trading

[ShaunWhite]

I think you'd have to be a pretty special trader to find something Peter & Dallas weren't already aware of.

I don't think it's especially groundbreaking, they'll have been there I'm sure. I just want to see what they did with it or similar.

That's why I'm biased against execution psychology Shaun

I do see why. Days when I'm 'in the zone' I agree with you. Those days when I do something 'off piste' I don't. But obv I wouldn't go off piste if I was certain about my strategies, check mate....so you win I think, 2-1 AET

I'm looking for help just as the next man; & love talking about the hard quantitative aspects of trading

When's that 'Win a Day with...' comp being run again?
I flunked maths A-level because I wasn't concentrating when they did differentiation, everything after that used it and I was screwed. Now I understand it it's straightforward. This feels like my differentiation moment all over again.

[ShaunWhite]

I've found what I think is a ground-breaking result on certain markets; but the ramifications of the discovery means that I cannot trade certain (what many people would call) "bread-&-butter" markets.

Do you mean the sort of scenarios that clash?

Like finding strategies for tennis games, sets and matches and then finding it hard to visualise the game strategy while it's wrapped inside the other two. Or how perhaps scalping whist having a swing strategy?

[ruthlessimon]

Do you mean the sort of scenarios that clash?

It's literally a clash of the "two forces". Mean reversion & momentum.

My automated statistics dictate that there's a bias for one force to predominate given x conditions - but - I know top manual traders; are about to take the opposite of my trade - & win. Therefore it means, I'm missing something, even though it's profitable longterm to go against them. That leads to a humungous mental block, to such an extent that I struggle to even execute. Technically that could be labeled as an execution error (i.e. aweful psychology ).

I've got two choices:

1. I accept the edge as I've discovered it - even though I'll be reminded daily there's potential to improve it. Work on my execution pschyology & just execute like robot
2. Work out how to improve the edge, & solve why there are traders going against me, & winning longterm (i.e. more profitable than me - I'm a competitive guy )

I go for the latter, because that'll see me improve hugely; & could be a step to further discovery. Option 1, is a regression & negative IMHO.

[ruthlessimon]

Also, option 1 means being at a huge risk to change. A market shift, sees it erode.

By having extra layers of complexity hedges that risk

[Lucacrebbe]

Since there was quite a bit of interest in relation to the process of strategy development (mainly modelling), I thought it was certainly worthy of its own thread. & if this receives continued interest, could become a focal point of future BA education.

I have a feeling traders are simply unaware of when to correctly use a ‘modelling’ or ‘backtesting’ methodology (when testing a strategy; especially pre-race traders). This mundane, overlooked problem, is actually a key reason for a lack of consistency & poor returns for a trader.

Shaun provided a great example in a previous thread.

A trader unaware of the results of previous shortie races, will miss vital context going into the next races. This is something that is missed when backtesting. However, a trader who understands ‘modelling’, will be able to spot the pattern, act faster & with more confidence.

To conclude, I think the majority of traders are underdeveloped in the technique of modelling. And more information on the topic could go a long way to getting users who are negative, positive. And those who are positive, exponential.

If you’re interested in learning more about modelling or have any thoughts on the technique, drop a post

I am interstead I would like to know more about it

[ShaunWhite]

This one has bugged me since it was written about 3 months ago

So I then started to look at why the large moves occurred.

Challenge accepted.

Simon, I thought I'd post this below so you can see why I'm invogorated about this topic again.

Orange - Price
Blue - Direction signal generated in real time.

Every one just before the move and it's nothing to do with wom or even anything I've seen mentioned before....and I've really looked.

Obviously they're not all like that or I'd have a much better tan. Seems to be a case now of just finding out why it's worked on the ones it has and visa versa. I've got too many of these to be a coincidence but the usual filters are either irrelevent or ineffective. Not enough hours in the day atm.

[Wolf1877]

Interesting Shaun. I just had a look at my data for that race. Not sure if you have obfuscated or you have a bug. I did find the matching data though!

[LinusP]

By having extra layers of complexity hedges that risk

I find the opposite is true.