Glad to hear my work helped you out. If you ever want more than what I offer free, check out my workshop: www.kjtradingsystems.com/strategy-workshop.html
I’ve been using a combination of a max % at risk, and sizing contracts based on buying power available. I check both, and make sure that neither is outside my range of acceptability. Sometimes I could go with more contracts based on a max risk of 2+/-%, but that would use more than 50% of my Buying Power. Those are my limits. So my risk is greater than 2+/-% or I’m using more than 50% BP, I’ll dial down the number of contracts so those are both in line. That’s served me pretty well. It could be that I could increase the % of BP required, and stay within the 2% risk. I’ll take a look at that. I hate losses, and sizing according to the Martingale system or Kelly Criterion scare me! I laughed out loud at the comment, “...the worst you can do is blow out your account!” Good video. I want to look into this more!
Awesome video!! When you do the backtest do you always assume one contract every time? If you test from 2006 to 2020 using the walk-forward method it always would be one contract or do you add position sizing methods? Also, do you keep the slippage and commissions fix in all periods of your backtest? Thank you!
Yes I do, for both questions. I find it easier to apply position sizing after developing the base strategy. Commissions and slippage have not changed a lot the past 10 years (in the futures markets), so I keep those constant. THANKS!
As always, very useful video Kevin! I have a qn regd % risk and whether it makes sense to apply it when trading a basket of stocks. Let's say I am trading a universe of 200 stocks. I have a condition that I will not enter into more than 10 open positions at any time. So, if a 11th entry signal comes in (while there are already 10 open positions), it gets dropped. Now consider if I use say a 3% risk PS strategy. My starting capital is say 10000 (dollars, pounds, whatever). My 3% risk says I can risk 300 on the first trade. I get an entry signal for stock A, which is priced at 1000 and the stop-loss is say 50. I can thus buy 6 shares (300 / 50) of this stock. This costs me 6000 (6 * 1000). The remaining capital is thus only 4000 (10000 - 6000). Now, I get an entry signal for stock B, for which I will take 3% of the remaining capital (4000) and so forth. If you observe, the capital remaining for each successive stock dramatically reduces (in fact, you will get something like an exponentially decaying curve), until one or more exits happen, freeing up capital. This being the case, the results become very sensitive to the order of entries. The amount of money used to purchase a stock becomes inversely proportional to the current number of already open positions. So, the equity curve for entries happening in the order A, B, C, D, etc will be very different from the curve for say C, D, A, B, etc (In the former case, stock A will get a huge size. In the latter, stock C will get a huge size). The final result thus seems to be more due to random chance (the order in which stocks enter). What are your thoughts on this? Is there a more sensible way to PS when trading a number of instruments (stocks here) that use the same capital? (Apologies for the lengthy comment, but I wanted to give a concrete example).
You are now tying the performance of all those trades together, since size of one depends on size of previous. I would try to make each trade as independent as possible from all other trades.
@@AlgoTradingWithKevinDavey By that do you mean a position sizing approach that doesn't cause this dependence? So instead of percent risk, maybe something like allocating one tenth of the capital to a new entry? That sounds like there won't be any preferential treatment.
Hi! Great content and am reading your book at the moment on building algorithmic systems, thank you for that. But i kind of don't get the Kelly results since it should be the optimal amount statistically. I see percentages in the table legend like 14.1%, 19.87% and so on, but cant seam to figure out what are those numbers. As i understand Kelly it should go like this: 1% of total Equity / Average Loss (Points/Pips or Stop loss if is set) = Position size or do get something wrong here. Since if you loose then downsize position proportionally and as you gain Equity increase position size proportionally as well. 1% is kinda optimal but still high risk, statistically should work up to 2%, but from 0-1% seams that the most bang for the buck is at.
Thanks for reading my first book! What you are seeing with Kelly is 1) the impact of randomness and 2) the constraint of real world (limited money) effects. When you look at approaches like Kelly or Optimal f, the theory is one thing, and actual practice is another. Both methods can be really dangerous...
Great video! Was looking forward to it. Thank you very much. Question: What do you think about sizing based on where the stock is in it's 52-week high/low range or trailing N-day high / low range? This closely aligns with retracement based entry/exit. I'm going to start looking into defining the contract size for range of retracement. Eg. if the price has retraced more than 20% and has a statistical edge at the level of retracement then trade more contracts versus a retracement at a 8% level even if there is statistical evidence of a smaller pullback. Would love your thoughts on it.
Hi, great video as always. When looking at the Gold equity strategy at 13:29. It doesnt make any sense for me on how the equity grow from 200k to 1million. Based on the rules, double the lots for each lots. This means that the net gain for each recovery/higher equity watermark should be 1 lot. Example, losing 1, 2, 4, then 8(win) . In net it is 8-4-2-1=1 lot gain of profit. Technically, the equity wouldnt jump from 200k to 600k immediately? It should be growing steadily and slowly instead. Or I am missing something over here?
Using risk based Martingale distributed over high probability range. Add to full size if reversal/trend is confirmed. Stop out at 1-1.5R. This approach keeps drawdowns very low.
Hello. Great video. I've heard some people use fractional Kelly as the position size method, this way it's harder to blow their account and the trade the largest when they are on a winning streak and trade the smallest on a losing streak. What do you think? I think it's a great method.
Would I be correct in saying that this doesn't account for equity needed to buy the contracts and the return % is not adjusted for that? For example, when the martingale dips to 200k, it assumes there is enough in the account to maintain 2^x contracts at that moment... Or is that negligible in this case? At $11,000 a pop for gold, it seems significant and would mean "starting equity" might not be accurate for martingale depending on how early your dip occurs... whereas adding 1 contract every 25k or 50k of profit would mean "starting equity" is probably truly all you need. In summary, I'm saying the "buffer" needed to support high risk strategies to ensure there is always enough in the account to maintain/increase size upon dips is not accounted for in the P/L%, and hence the "starting equity" isn't necessarily the same for all the strategies. This can skew the less risky methods to be actually much better in terms of return %. Does that make sense?
Excellent points! I did not include initial margin in this analysis, so you have a valid point. I probably will do a follow up video to this one soon, since many people are asking about some different position sizing approaches. I'll make sure to cover the impact of initial margin requirements in that video. Thanks for the suggestion, I definitely appreciate it!
Interesting analysis. In Dr Howard Bandy's book Quantitative Technical Analysis, he recommends doing an MC simulation of the strategy, and using a PS algorithm based off the user's max drawdown tolerance of the 20th percentile of the MC using a rolling window. Easier to use for stocks & forex which scale better than futures. Have you ever tested something like this? (But probably this PS algo is less relevant for you since you trade futures). I starting coding it up in python awhile ago but got side-tracked. Thanks for this video!
I've talked to Dr. Bandy a few times, but we never specifically spoke about this.Certainly worth testing, and with the micro mini stock index futures, it may be very good to do.
Another great video for my Davey Library:) After taking your courses, I would say I have a propensity towards the potentially lower risk of diversification over position sizing. At what point does diversification of proven non-correlated systems have no greater effect on the total account equity curve and or MAR? Also, what does the “J” in KJ Trading stand for?
Glad you enjoyed it! That question is really hard to say, but I would guess 20-30 strategies in different markets is a good place to be. "J" is for my middle name...I'll leave you to guess what exactly (think Beatles song)...
Thanks for very informative video Kevin!! In my current futures trading system, I am using PS based on the volatility of the bar with 3% risk of equity as my stop loss is based on the volatility (ATR). So if the volatility is less then it will buy more contract and if is more (what we have seen in March'20) then it will buy/short less contracts. Also in a way it has built in compounding mechanism. With increase in your equity your trade qty will increase and will decrease with decrease in equity. Do you see any drawback in this approach?
Thanks for the question. I have never really tested that exact approach - I've done some things close to it - so you'll want to evaluate it and see if you like the return and drawdown characteristics.
Martingale dont maje sense, if you flip a coin and have 3 tails, dont mean the problability of next flip is small of gettig tail, its 50% equal any flip you made... The same aproach as martigale is double your bet randonly for no reason...
![039: Michael Cook discusses position sizing, stops and discretion in algo trading. [AUDIO ONLY]](http://i.ytimg.com/vi/H2uIIoyb44U/mqdefault.jpg)
![Megan Thee Stallion - Neva Play (feat. RM) [Official Video]](http://i.ytimg.com/vi/TpYTyAaTRts/mqdefault.jpg)
Grab Kevin's free mini S&P algo strategy and other helpful trading info: kjtradingsystems.com/freeyoutube21.html
Excellent content! Really helped me understand the complexity involved in position sizing. So many ways to go about it! Keep them coming !!!
Thanks JB, will do!
Dear Davey, it was a good video, all of your books especially “building algo systems” is a gem. I inspired a lot from your books. Thanks again
Glad to hear my work helped you out. If you ever want more than what I offer free, check out my workshop: www.kjtradingsystems.com/strategy-workshop.html
12:25 "Nerves of steel or some other body part" Hahahaha
Glad you got the joke!!!
Nice video as usual! If you have the time a video about portfolio aspects of position sizing would be really intersting.
Thanks, I will put it on the list.
I’ve been using a combination of a max % at risk, and sizing contracts based on buying power available. I check both, and make sure that neither is outside my range of acceptability. Sometimes I could go with more contracts based on a max risk of 2+/-%, but that would use more than 50% of my Buying Power. Those are my limits. So my risk is greater than 2+/-% or I’m using more than 50% BP, I’ll dial down the number of contracts so those are both in line. That’s served me pretty well. It could be that I could increase the % of BP required, and stay within the 2% risk. I’ll take a look at that. I hate losses, and sizing according to the Martingale system or Kelly Criterion scare me! I laughed out loud at the comment, “...the worst you can do is blow out your account!” Good video. I want to look into this more!
Thanks for the comment, I'm glad this video got you thinking!
Thanks for video. looking forward for video about correlation and position sizing methods for portfolio of trading systems.
Thanks, it might be a while, that is definitely a very involved subject.
Great videos as always!! Would love to get your opinion on fractional kelly or a video. Right now I use quarter kelly.
I personally don't use Kelly. Even 1/4 Kelly can be high risk, depending on strategy.
Very educational thank you
Glad you enjoyed it
Awesome video!! When you do the backtest do you always assume one contract every time? If you test from 2006 to 2020 using the walk-forward method it always would be one contract or do you add position sizing methods? Also, do you keep the slippage and commissions fix in all periods of your backtest? Thank you!
Yes I do, for both questions. I find it easier to apply position sizing after developing the base strategy. Commissions and slippage have not changed a lot the past 10 years (in the futures markets), so I keep those constant. THANKS!
@@AlgoTradingWithKevinDavey Thanks!!!
Great stuff, Kevin! At least I will watch out for use the Martingale MM as it was suggested in one course I've recently bought :D Hi, from Russia :)
Martingale almost always leads to disaster...
As always, very useful video Kevin! I have a qn regd % risk and whether it makes sense to apply it when trading a basket of stocks.
Let's say I am trading a universe of 200 stocks. I have a condition that I will not enter into more than 10 open positions at any time. So, if a 11th entry signal comes in (while there are already 10 open positions), it gets dropped. Now consider if I use say a 3% risk PS strategy. My starting capital is say 10000 (dollars, pounds, whatever). My 3% risk says I can risk 300 on the first trade. I get an entry signal for stock A, which is priced at 1000 and the stop-loss is say 50. I can thus buy 6 shares (300 / 50) of this stock. This costs me 6000 (6 * 1000). The remaining capital is thus only 4000 (10000 - 6000). Now, I get an entry signal for stock B, for which I will take 3% of the remaining capital (4000) and so forth.
If you observe, the capital remaining for each successive stock dramatically reduces (in fact, you will get something like an exponentially decaying curve), until one or more exits happen, freeing up capital. This being the case, the results become very sensitive to the order of entries. The amount of money used to purchase a stock becomes inversely proportional to the current number of already open positions. So, the equity curve for entries happening in the order A, B, C, D, etc will be very different from the curve for say C, D, A, B, etc (In the former case, stock A will get a huge size. In the latter, stock C will get a huge size). The final result thus seems to be more due to random chance (the order in which stocks enter).
What are your thoughts on this? Is there a more sensible way to PS when trading a number of instruments (stocks here) that use the same capital? (Apologies for the lengthy comment, but I wanted to give a concrete example).
You are now tying the performance of all those trades together, since size of one depends on size of previous. I would try to make each trade as independent as possible from all other trades.
@@AlgoTradingWithKevinDavey By that do you mean a position sizing approach that doesn't cause this dependence? So instead of percent risk, maybe something like allocating one tenth of the capital to a new entry? That sounds like there won't be any preferential treatment.
Hi! Great content and am reading your book at the moment on building algorithmic systems, thank you for that. But i kind of don't get the Kelly results since it should be the optimal amount statistically. I see percentages in the table legend like 14.1%, 19.87% and so on, but cant seam to figure out what are those numbers. As i understand Kelly it should go like this: 1% of total Equity / Average Loss (Points/Pips or Stop loss if is set) = Position size or do get something wrong here. Since if you loose then downsize position proportionally and as you gain Equity increase position size proportionally as well. 1% is kinda optimal but still high risk, statistically should work up to 2%, but from 0-1% seams that the most bang for the buck is at.
Thanks for reading my first book! What you are seeing with Kelly is 1) the impact of randomness and 2) the constraint of real world (limited money) effects. When you look at approaches like Kelly or Optimal f, the theory is one thing, and actual practice is another. Both methods can be really dangerous...
Great video! Was looking forward to it. Thank you very much.
Question: What do you think about sizing based on where the stock is in it's 52-week high/low range or trailing N-day high / low range? This closely aligns with retracement based entry/exit.
I'm going to start looking into defining the contract size for range of retracement. Eg. if the price has retraced more than 20% and has a statistical edge at the level of retracement then trade more contracts versus a retracement at a 8% level even if there is statistical evidence of a smaller pullback.
Would love your thoughts on it.
I have never tried it. It definitely warrants attention and testing!
Hi, great video as always. When looking at the Gold equity strategy at 13:29. It doesnt make any sense for me on how the equity grow from 200k to 1million.
Based on the rules, double the lots for each lots. This means that the net gain for each recovery/higher equity watermark should be 1 lot.
Example, losing 1, 2, 4, then 8(win) . In net it is 8-4-2-1=1 lot gain of profit.
Technically, the equity wouldnt jump from 200k to 600k immediately? It should be growing steadily and slowly instead.
Or I am missing something over here?
Thanks for the comment, I'll have to look at those results more closely to answer the question. Feel free to send me an e-mail.
Using risk based Martingale distributed over high probability range. Add to full size if reversal/trend is confirmed. Stop out at 1-1.5R. This approach keeps drawdowns very low.
Can you e-mail me at kdavey@kjtradingsystems.com? I'll try this out, but I have a few questions to get it right.
How do you do this?
Hello. Great video. I've heard some people use fractional Kelly as the position size method, this way it's harder to blow their account and the trade the largest when they are on a winning streak and trade the smallest on a losing streak. What do you think? I think it's a great method.
Thanks for the info! I don't personally use this, but it is worth investigating.
How do you calc fractional Kelly when trading?
Would I be correct in saying that this doesn't account for equity needed to buy the contracts and the return % is not adjusted for that?
For example, when the martingale dips to 200k, it assumes there is enough in the account to maintain 2^x contracts at that moment... Or is that negligible in this case? At $11,000 a pop for gold, it seems significant and would mean "starting equity" might not be accurate for martingale depending on how early your dip occurs... whereas adding 1 contract every 25k or 50k of profit would mean "starting equity" is probably truly all you need.
In summary, I'm saying the "buffer" needed to support high risk strategies to ensure there is always enough in the account to maintain/increase size upon dips is not accounted for in the P/L%, and hence the "starting equity" isn't necessarily the same for all the strategies. This can skew the less risky methods to be actually much better in terms of return %. Does that make sense?
Excellent points! I did not include initial margin in this analysis, so you have a valid point. I probably will do a follow up video to this one soon, since many people are asking about some different position sizing approaches. I'll make sure to cover the impact of initial margin requirements in that video. Thanks for the suggestion, I definitely appreciate it!
Interesting analysis. In Dr Howard Bandy's book Quantitative Technical Analysis, he recommends doing an MC simulation of the strategy, and using a PS algorithm based off the user's max drawdown tolerance of the 20th percentile of the MC using a rolling window. Easier to use for stocks & forex which scale better than futures. Have you ever tested something like this? (But probably this PS algo is less relevant for you since you trade futures). I starting coding it up in python awhile ago but got side-tracked. Thanks for this video!
I've talked to Dr. Bandy a few times, but we never specifically spoke about this.Certainly worth testing, and with the micro mini stock index futures, it may be very good to do.
Another great video for my Davey Library:) After taking your courses, I would say I have a propensity towards the potentially lower risk of diversification over position sizing. At what point does diversification of proven non-correlated systems have no greater effect on the total account equity curve and or MAR? Also, what does the “J” in KJ Trading stand for?
Glad you enjoyed it! That question is really hard to say, but I would guess 20-30 strategies in different markets is a good place to be. "J" is for my middle name...I'll leave you to guess what exactly (think Beatles song)...
@@AlgoTradingWithKevinDavey Thank you for the input, Im going to guess "Jude"?
although the song came out after I was born!
@@AlgoTradingWithKevinDavey Perhaps you were the unknown inspiration:)
Thanks for very informative video Kevin!! In my current futures trading system, I am using PS based on the volatility of the bar with 3% risk of equity as my stop loss is based on the volatility (ATR). So if the volatility is less then it will buy more contract and if is more (what we have seen in March'20) then it will buy/short less contracts. Also in a way it has built in compounding mechanism. With increase in your equity your trade qty will increase and will decrease with decrease in equity. Do you see any drawback in this approach?
Thanks for the question. I have never really tested that exact approach - I've done some things close to it - so you'll want to evaluate it and see if you like the return and drawdown characteristics.
@@AlgoTradingWithKevinDavey Thank you for the reply, sure I will evaluate that
fixed ratio sizing
I'll include it if/when I create followup video. People have blown out accounts with fixed ratio though.
Which PS do you most use Kevin?
I tend to stay away from the really risky ones, and I use a few different ones (the simpler ones I show here), depending on what I am trying to do.
Martingale dont maje sense, if you flip a coin and have 3 tails, dont mean the problability of next flip is small of gettig tail, its 50% equal any flip you made... The same aproach as martigale is double your bet randonly for no reason...
Thanks for the comment, many people also use an anti-martingale approach. Both can be dangerous.
@@AlgoTradingWithKevinDavey what is the anti-Martingale approach?
@@William.-. Thanks for comment. This explains it better than I ever could: www.investopedia.com/terms/a/antimartingale.asp
@@AlgoTradingWithKevinDavey ty for the quick respond. Didn't realize that TH-cam blocked people from clicking links on comment
@@William.-. - weird. It is top rank link when you google anti-martingale
king davey
Not king of much, I am afraid, but thanks anyhow!!
It's not a martingale
Can you explain?