risk:reward and Kelly's criterion- Holy schmoley !

[Scotty2Cues]

risk:reward and Kelly's criterion

Been reading about Kelly's criterion. It gives the fraction of your account you should bet to give you the most optimal account growth if you know the probabilities of a win/loss.

The formula is f = [p(R+1)-1]/r

f = fraction to risk,
p = probability of win
R = reward as in 1: R

Then on average, each trade would grow the account by

y = [(1+Rf)^p][(1-f)^(1-p)]

Assume we have a trade setup with 60% probability of winning and our risk:reward is 1:1

Kelly's formula gives f = 20%!! and y = 1.020 so after 10 trades we can expect our starting account to have grown by 22%

Or if we risk 3% then after 10 trades we can expect account to have grown by 5.7%

(excellent article on Kellys criterion by Tom Weideman here http://groups.google.com/group/rec.gambling.poker/msg/7bb09884cfac7678

 

[Lord Flasheart]

Been reading about Kelly's criterion. It gives the fraction of your account you should bet to give you the most optimal account growth if you know the probabilities of a win/loss.

The formula is f = [p(R+1)-1]/r f = fraction to risk,
p = probability of win
R = reward as in 1: R

Then on average, each trade would grow the account by

y = [(1+Rf)^p][(1-f)^(1-p)]

Assume we have a trade setup with 51% probability of winning and our risk:reward is 1:1

Kelly's formula gives f = 2% and y = 1.0002 so after 10 trades we can expect our starting account to have grown by 0.2% [1.0002^10 = 1.002]

Now consider a setup with 51% probability of winning but with risk:reward 1:2

Now f = 26% !! and y = 1.07 so after 10 trades our original account has almost doubled!

If we risk 2% on risk:reward 1:R and 51% success y = 1.01 so our account increases by 10%

(excellent article on Kellys criterion by Tom Weideman here http://groups.google.com/group/rec.gambling.poker/msg/7bb09884cfac7678

Looks a little too complicated for me. I dont understand when people talk about the amount they risk from their account. I often risk 10% or more. Why? simple, i never keep much in my account. I keep most in my bank account,so does that count as my trading account.I transfer accross as needed so the whole not is not at risk to a spike etc. If i was taken out say like sept 11 with a huge gap and fall. i could and would quite happily knock them,i certainly wouldnt let them take family assets etc. so why keep 100% in your account to trade 1%. No sense at all to me.

 

[Scotty2Cues]

Looks a little too complicated for me. I dont understand when people talk about the amount they risk from their account. I often risk 10% or more. Why? simple, i never keep much in my account. I keep most in my bank account,so does that count as my trading account.I transfer accross as needed so the whole not is not at risk to a spike etc. If i was taken out say like sept 11 with a huge gap and fall. i could and would quite happily knock them,i certainly wouldnt let them take family assets etc. so why keep 100% in your account to trade 1%. No sense at all to me.

I guess if you are taking money from your bank account, then its counts as your trading account, or more like what you are prepared to risk...

 

[Lord Flasheart]

its difficult isnt it. What if you have £20,000 in your savings account and £10 k in your current account and 5k in shares,is that your trading capital. Thats why i look more at r/r than %. i guess its the amount you say you are prepared to lose over time,but then you need to put in more to get the leverage,so does the levereaged amount not then count, etc,etc

 

[Scotty2Cues]

its difficult isnt it. What if you have £20,000 in your savings account and £10 k in your current account and 5k in shares,is that your trading capital. Thats why i look more at r/r than %. i guess its the amount you say you are prepared to lose over time,but then you need to put in more to get the leverage,so does the levereaged amount not then count, etc,etc

Ignoring the £5k in shares, if you are prepared to lose the £20k and £10k on trading then your trading account is £30k and if you are prepared to lose £500 on a trade then youre risking 1.7%

If your'e prepared to lose only 10 of the £20k then your trading account is £20k and risking £500 on a trade is 2.5% of your account.

Dont think you count the £5k as its being risked and the equity changes, but I suppose if the £5k you risked was in profit and you had a stop to brekeven, that gives you at least another £5k on top of your other accounts

 

[JRP2891]

Re: risk:reward and Kelly's criterion

Been reading about Kelly's criterion. It gives the fraction of your account you should bet to give you the most optimal account growth if you know the probabilities of a win/loss.

The formula is f = [p(R+1)-1]/r

f = fraction to risk,
p = probability of win
R = reward as in 1: R

Then on average, each trade would grow the account by

y = [(1+Rf)^p][(1-f)^(1-p)]

Assume we have a trade setup with 60% probability of winning and our risk:reward is 1:1

Kelly's formula gives f = 10%!! and y = 1.015 so after 10 trades we can expect our starting account to have grown by 16%

Or if we risk 3% then after 10 trades we can expect account to have grown by 5.7%

(excellent article on Kellys criterion by Tom Weideman here http://groups.google.com/group/rec.gambling.poker/msg/7bb09884cfac7678

I've looked at Kelly in the past, but ultimately your money management comes down to how much pain you can take without it affecting how you trade. It's nice to be able to sleep at night, and if you're using the unadulterated Kelly criterion you'll soon be an insomniac.

 

[JRP2891]

I keep most in my bank account,so does that count as my trading account.I transfer accross as needed so the whole not is not at risk to a spike etc. If i was taken out say like sept 11 with a huge gap and fall. i could and would quite happily knock them,i certainly wouldnt let them take family assets etc. so why keep 100% in your account to trade 1%. No sense at all to me.

Surely you're still liable for any losses you incurred regardless of how much you have in your trading account?

 

[meanreversion]

I read an article on Kelly (on the internet, so it must be true), where the idea was that you calculate the Kelly ratio and then scale it by the amount you're prepared to lose (you could lose all your money betting raw Kelly).

So if the ratio is 20% (i.e. bet 20% of your account), but you're only prepared to lose 20% of your account, then bet 4% per trade (20% of 20%).

I still prefer backtesting, looking at CAR/DD, Sharpe ratio, that type of thing. They'll probably come out with similar numbers any way.

 

[JRP2891]

I read an article on Kelly (on the internet, so it must be true), where the idea was that you calculate the Kelly ratio and then scale it by the amount you're prepared to lose (you could lose all your money betting raw Kelly).

So if the ratio is 20% (i.e. bet 20% of your account), but you're only prepared to lose 20% of your account, then bet 4% per trade (20% of 20%).

I still prefer backtesting, looking at CAR/DD, Sharpe ratio, that type of thing. They'll probably come out with similar numbers any way.

That sounds a more palatable version. Most traders who use a watered down Kelly can't get near half the amount suggested, but it could still be handy for being precise in your money management if you know your exact trade results.

 

[Scotty2Cues]

I suppose you could use it by being conservative with probability assignments. Lets say you have a setup with r:R of 1:1 that you have 80% confidence in,

Kellys formula gives f = 60% with account grown to 6 times its size after 10 trades


We could assume a win rate of 55% and go from there, mind you that gives f = 10% with account increased by 5% after 10 trades...

 

[wackypete2]

Why make things so difficult with unnecessary math calculations? Keep it simple. If I am prepared to accept a 1% loss, then risk 1% of my account, or whatever is your threshold. Too much time spent on calculating R, kelly, sharp,.etc....will only make you lose focus. The only thing that's important is are you profitable without stressing.
I do some simple calculations at the end of the day, but can't be bothered while trading!

Peter

 

[meanreversion]

It's important to spend as much time as possible calculating obscure ratios. Don't be a LUDDITE !

 

[Scotty2Cues]

I suppose you could use it by being conservative with probability assignments. Lets say you have a setup with r:R of 1:1 that you have 80% confidence in,

Kellys formula gives f = 60% with account grown to 6 times its size after 10 trades


We could assume a win rate of 55% and go from there, mind you that gives f = 10% with account increased by 5% after 10 trades...

Say another setup gives 1:1.25 (we look for another 10pips reward on our 40 pip : 40pip r:R) and assume 55% setup, Kelly gives f = 19% with account growth of 25% after 10 trades

 

[Scotty2Cues]

Why make things so difficult with unnecessary math calculations? Keep it simple.
Peter

this helps

 

[scose-no-doubt]

whats maths got to do with crossing the roads?

*ba dum psh*

 

[Scotty2Cues]

whats maths got to do with crossing the roads?

*ba dum psh*

good one!

 

[wackypete2]

I was clicking on it. I thought I was voting for Scotty in the t2w awards.

Peter

 

[meanreversion]

Isn't the Green Cross Code man the same chap who played Darth Vader in Star Wars?

 

[Lord Flasheart]

Surely you're still liable for any losses you incurred regardless of how much you have in your trading account?

yes you may be, but before i considered paying out, id want to know that the market never touched my price by any market maker etc.I would find any way of contesting the amount owed. Would you sell your house to pay a gambling debt?

 

[robster970]

Isn't the Green Cross Code man the same chap who played Darth Vader in Star Wars?

Dave Prowse I think he was called. Was only in the first one as well (Star Wars IV or whatever horsesh1t george lucas concocted). Anyway the first one - you know.