Meritocratic System Reward

pkfryer

Active member
Messages
243
Likes
0
I have been having this conservative idea in the back of my mind for a while now but Im not sure how useful it would be.

When confronted with a new system that you have either backtested yourself or seen the past results of, it is still no guarantee that it will perform well in the future.

So I was wondering whether an approach of meritocratic based trading size adjustment would reduce risk whilst allowing a system or method to show its true colours. In other words, you reward a system for every point it gives in profit and punish a system for every loss. Anything above a certain target level of profit (e.g. 100 points) is given a maximum of 25% increase in trade size for the next trade. E.g. if you get half the target profit you can reward it 12.5% increase in trade size. The target profit must not be adjusted frequently. Also, you can only increase the trade size if it doesn't break your risk/money management rules!

For any loss it would be an automatic reducation of 5%. I chose 5% as I know that some systems give good signals less than 50% of the time but the profitability factor still makes it an overall winner. This would prevent a good system from being continually reduced but it would stalemate a system that was consistently failing to achieve its target profit per trade. This would either mean the system isn't very good, or you are expecting too much of it! A losing strategy would get wittled away to nothing in a very short period of time.

I haven't got any back testing software that could test this strategy out. So Im wondering if anyone would like to see if it would do a good job against protecting you from a dodgy system and letting you milk a good system for all its worth! Would it be a good strategy? What would be the results of changing the percentage adjustment and the profit target?

My only concern with this is that it might make it harder for profitable trades to counterbalance the bad trades. A string of 4 losses would reduce a compounded 20% of trading size. However, a string of only 2 good wins would swell the trade size by 50% (compounded), increasing the overall trading size even after a few losses.

I think this approach would also protect us from a system thats slowly going out of synch with the market. But it would be interesting to backtest this approach and see if it does a good job.

Maybe starting off really conservatively at 100 shares (or less) depending on your account balance. I wonder how many shares this would increase to and how fast, with some previously backtested systems?
 
Hi pkfryer,

Interesting idea.
However, isnt this the similar as saying you are running several trading styles in parallel ?
Suppose you cannot decide on a strategy.
You decide to spread your risk by running several strategies in parallel.
Suppose StrategyA is based on MAs.
StrategyB is based on Stochastics.
StrategyC is based on MACD.

You have £1000 in each strategy.
You run the strategies separately and in parallel.
You risk a percentage of the pot on each trade.

The strategy that wins increases in value, and thus has an increasing amount to trade next trade.
The strategy that loses, decreases in value, and thus has less to trade next time.

Is your idea just a variation of this ?

The main headache with this is that the strategies may contradict each other based on their parameters,
and can place psychological pressure on you, as you are placing contradicting trades.

As I say, an interesting idea.
 
I have used Metastock for backtesting. There is no facility for altering size during a test.
Maybe Tradestation does it ? Otherwise it's probably down to Excel.

A meritocratic alternative which appears to help is to plot a moving average on the Equity chart. (say 30 periods).
Once the equity value falls below the MA you stop trading the system. When it goes above, you start trading it again.
Somewhat rough and ready, but maybe good enough.

Glenn
 
reward/punishment system

It would be very easy for you to apply your scenario to the 5 day rsi system from www.vtoreport.com

You can cut and paste their trade list into excel and then make yourself an additional column or 2 to run your dollar totals and win/lose adjustments.

I did a variant of this by starting with 10,000 and letting it roll continuously from trade to trade, win or lose.

The result I got was 10 to 45 k since 1997 including all losing trades.

I continue to be amazed these guys sell nothing, offer the worlds simplest system, it works and is almost virtually unknown.

You can go to clearstation and see folks doing dozens of trades per month while vto handily beats all that work.
 
I have read in many books that the simplest most robust and de-optimised systems are the best! Complexity doesn't necessarily mean greater profits. I was shocked that there system only seems to trade once a month!! But with such a good win probability its worth following, even as just a little trading side line! I do that with a few advisory services... when they give advice I take it and set up a trade, and get on with my own methods along side!

I think I'll start observing their recommendations!

I'm going to learn Excel so I can see what effect the meritocratic approach would have on equity!
 
The type of strategy you are describing is basically an equity curve stategy and the Rina add on for Tradestation will let you test these.

If you apply this type of strategy mechnically to trend following systems it doesn't tend to work that well as it reduces position size to a minimum just before a really big trend. This kills the overall profitability of the system.

A good money manager will sometimes adjust position size of a system in a portfolio if it is starting to produce results that are not consitent with history. However, detemining if the system is starting to fail is difficult and requires experience and discretion,
 
Thanks reeve.. this is an approach that is out there already? I've never heard of it before. But then, I suppose there is nothing new under the sun!

That is my thought too, that after a losing streak you would have your trade size reduced, making it harder for the systems to work! For instance, if you have a system with a low win probability but a high profitability ratio you'll be in trouble. E.g. a string of 5 losses would decimate the trading size by 25%, even if you hit a real big win that makes you overall profitable, the system would eventually wittle the size of the shares down if used mechanically! An alternative could be to use the largest loss and largest win and use this to determine increase or decrement of trade size!

50% allocated to the largest win, 50% to largest loss... so the closer your win is in comparison to the largest win the more it will go up, and the smaller the loss the less it will go down.

I do believe that this approach has potential even if it needs refinement! We could simultaneously run several systems and when one starts to outperform the other the meritocractic system would begin to lean towards the winning formula whilst reducing the losing systems size. Having a mechanical method of rewarding systems is prefered to subjective decision making! Like the common beginners flaw of becoming attached to a trade, we could become emotionally attached to a clapped out system!
 
Every hedge fund and CTA has been faced with the type of problem you are looking at and so a great deal of time and effort (and not a little money) has been spent looking at every conceivable form of money management strategy including strategies of the type you have described.

I tend to have multiple systems with some trading real money and some waiting in the wings paper trading. If one system starts to show poor performance I will start to fade it out of the portfolio whilst starting to allocate some money to one of the paper trading systems if it is performing well. Over time the portfolio very slowly evolves using this method.
 
You guys may want to look into Optimal-f dynamic position sizing introduced by Ralph Vance (or Vince - cant remember). Do a search on google.

The idea is that you 'bet more' as your winning streak continues and decrease as you face drawdown. Optimal-f is the % of capital you risk per bet/trade.

Quite complex stuff (for me), but will appeal to those into mechanical systems.
 
Beware of optimal-f, f the in optimal f is the limit you can trade at which will blow your account.
It is not really optimal and so the people that came along after wards came up with "safe-f" or "secure-f".

If you like money management chack out this site where you will be able to spend countless hours wading through the endless possibilities.

http://www.turtletradingsoftware.com/forum/viewforum.php?f=7
 
jmreeve said:
Beware of optimal-f, f the in optimal f is the limit you can trade at which will blow your account.
It is not really optimal and so the people that came along after wards came up with "safe-f" or "secure-f".

If you like money management chack out this site where you will be able to spend countless hours wading through the endless possibilities.

http://www.turtletradingsoftware.com/forum/viewforum.php?f=7

Alternatively, if you would like a (fairly) simple and lucid run down on how to define the optimal stake, see John Haigh: Taking Chances - winning with probability (Oxford University Press, paperback edition published 2000) which discusses essentially this problem in Ch 11 and Appendix V. This all goes back to the work of Kelly (you can find his original paper somewhere on this site, if you search under Kelly).
 
Top