So, Ive done some reading in the past few days, and obviously came across the prophet of bet trading, aka Ralph Vince. Im really impressed by his work, and i have learned so much, in this short period. You already know the story, but i have a few questions regarding Optimal f that i do not know the answers to. So here they are:
What is optimal f? As i get it, optimal f is a number, that we use to multiply our whole stake (the bank we use for the actual system), to get stake size for each of the individual bets, so we can maximize profit. Is that right? Okay, by that definition, i have "found" the optimal f for my system, but what now? How am i supposed to know what is the optimal f for the future systems? Do i just take this one and it doesnt change? It cant be, im sure that it does change with new bets. How do i use the optimal f in practice? Im really confused.
Btw, if youre wondering, im reading "The Handbook of Portfolio Mathematics Formulas for Optimal Allocation Leverage".
Thanks in advance
optimal f
- marko kostic
- Gaining experience
- Karma: 1
- arb12
- Totally Pro
- Karma: 23
Post
Re: optimal f
@Marko Kostic,
If you read Mr. Vince, you already know, that there are cases when Kelly's Optimal fraction and Vince's Optimal fraction are the same. There are also cases, especially in the trading, when Kelly's Optimal F is much bigger than 1.0 for instance, leveraged accounts, impractical for you. Naturally, a Kelly multiplier over 1.0 of your capital appears when the so-called by Vince "special case" for a given outcome appears, but much large capital for a given trade must be avoided. Fractional Kelly is a possible workaround here, but deeper estimations are needed. On these occasions, unlike Kelly, Vince's Optimal Fraction are between 0 and 1 of your capital.
In Vince's literature (books and papers) is also written a direct formula, the dependence between Optimal fraction accordin' to Mr. Vince, to the Mr. Kelly solution, even in dollar amount.
Optimal f = (Kelly Criterion Solution * -W) / S, where W is maximum loss.
Bear in mind, for Vince's application, historical data are needed. And your past performance is not equal to others' past performance, so some indirect assessments for your abilities to spot +EV may be applied and later help you in the analysis.
Naturally, updating and iterations are required. Backtesting also.
And after all, after reading them, if you trust these Authors, you must find the right implementation of the theory for your purposes.
Very, very short and clear represent of Vince's theory, plus some own ideas from the Author, please, read "Money Management Principles
for Mechanical Traders" circa 70 pages, see your e-mail.
Vince's book contains over 400, and much more in-depth maths inside, that's good.
In short, Geometric Optimisation in theory, when the advantage of your positions is real, proven +EV (not only computed one), when you've much big capital, in an endless row you'll have maximum gain versus other staking methods, but the pitfalls are very possible, i.e. much larger drawdowns, if bad losing streaks happen, causing you a leak of capital. Think about that possible situation: Run in Monte Carlo simulation several staking plans and probabilities. There is a possible situation - if your Geometric Optimisation 99.99 percent will give you much more gains over the long haul vs other selected by you staking methods, although you hold big +EV, the 100 000 simulations may return to you the probability of 0.01 percent to fail! Before reading these above carefully, do not stake for real! I've remembered, on one of the platforms circa 60 percent of the professional, experienced participants are wagering by flat staking method actually!
If you read Mr. Vince, you already know, that there are cases when Kelly's Optimal fraction and Vince's Optimal fraction are the same. There are also cases, especially in the trading, when Kelly's Optimal F is much bigger than 1.0 for instance, leveraged accounts, impractical for you. Naturally, a Kelly multiplier over 1.0 of your capital appears when the so-called by Vince "special case" for a given outcome appears, but much large capital for a given trade must be avoided. Fractional Kelly is a possible workaround here, but deeper estimations are needed. On these occasions, unlike Kelly, Vince's Optimal Fraction are between 0 and 1 of your capital.
In Vince's literature (books and papers) is also written a direct formula, the dependence between Optimal fraction accordin' to Mr. Vince, to the Mr. Kelly solution, even in dollar amount.
Optimal f = (Kelly Criterion Solution * -W) / S, where W is maximum loss.
Bear in mind, for Vince's application, historical data are needed. And your past performance is not equal to others' past performance, so some indirect assessments for your abilities to spot +EV may be applied and later help you in the analysis.
Naturally, updating and iterations are required. Backtesting also.
And after all, after reading them, if you trust these Authors, you must find the right implementation of the theory for your purposes.
Very, very short and clear represent of Vince's theory, plus some own ideas from the Author, please, read "Money Management Principles
for Mechanical Traders" circa 70 pages, see your e-mail.
Vince's book contains over 400, and much more in-depth maths inside, that's good.
In short, Geometric Optimisation in theory, when the advantage of your positions is real, proven +EV (not only computed one), when you've much big capital, in an endless row you'll have maximum gain versus other staking methods, but the pitfalls are very possible, i.e. much larger drawdowns, if bad losing streaks happen, causing you a leak of capital. Think about that possible situation: Run in Monte Carlo simulation several staking plans and probabilities. There is a possible situation - if your Geometric Optimisation 99.99 percent will give you much more gains over the long haul vs other selected by you staking methods, although you hold big +EV, the 100 000 simulations may return to you the probability of 0.01 percent to fail! Before reading these above carefully, do not stake for real! I've remembered, on one of the platforms circa 60 percent of the professional, experienced participants are wagering by flat staking method actually!