Mr Kelly - A betting game | Forex Factory

Options

Permalinks
Print Thread

Similar Threads

Kelly Criterion - Optimizing risk 42 replies

mastergunner's PipNailer & Kelly Criterian Journal 12 replies

Kelly EA 25 replies

Simple Derivation of Kelly's Formula 3 replies

Spread betting on options, good or bad? 0 replies

Attachments: Mr Kelly - A betting game
Exit Attachments

Mr Kelly - A betting game

  • #1
  • First Post: Jul 25, 2009 1:22am Jul 25, 2009 1:22am
Hi all,

I coded a very simple betting game called "Mr. Kelly" -- it's a demonstration of the effectiveness of the Kelly criterion for deciding how much to invest during a series of bets, which applies to forex and many other markets under simplifying conditions. You can find the file as an attachment below. Please let me know what you think about it!
Attached File(s)
File Type: zip mrKelly.zip   7 KB | 506 downloads
  • #2
  • Edited 1:47am Jul 25, 2009 1:33am | Edited 1:47am
faint this robot is too clever.

I took a 49% win ratio trade, robot never place a bet!!

probably that's what bucketshops is running upon... they sit there collect a spread and roll-over/swap and watch you place blind bets...
Attached File(s)
File Type: txt kelly_100.txt   10 KB | 443 downloads
  • #3
  • Jul 25, 2009 1:40am Jul 25, 2009 1:40am
Quoting billye
faint this robot is too clever.

I took a 49% win ratio trade, robot never place a bet!!

probably that's what bucketshops is running upon... they sit there collect a roll-over and swap...
Glad you liked it. "Mr. Kelly" will only bet if the expectancy (R-to-R*winProb - (1-winProb)) is positive, as the program explains to you every time Kelly wins.
  • #4
  • Jul 25, 2009 2:10am Jul 25, 2009 2:10am
Quoting outspan
Glad you liked it. "Mr. Kelly" will only bet if the expectancy (R-to-R*winProb - (1-winProb)) is positive, as the program explains to you every time Kelly wins.
i had a question, sorry for being noobs.. i set R to R is 4 win prob is 30% so according to equation, it is (4*30-(1-30)) = (120-(-29)) = 91 so mr kelly beter every round since this number is always same.. but how much is he betting? how did he calculate how much to bet
..
  • #5
  • Jul 25, 2009 2:18am Jul 25, 2009 2:18am
Quoting Holypendant
i had a question, sorry for being noobs.. i set R to R is 4 win prob is 30% so according to equation, it is (4*30-(1-30)) = (120-(-29)) = 91 so mr kelly beter every round since this number is always same.. but how much is he betting? how did he calculate how much to bet
..
The winning probability goes from 0 to 1 by definition, so the correct calculation is:

(4*0.3-(1-0.3))/4 = 12.5% of the current balance

In other words, divide the probability you entered, 30, by 100. I only made the range 1-99 so that it would be more convenient to type.
  • #6
  • Jul 25, 2009 2:43am Jul 25, 2009 2:43am
Quoting outspan
The winning probability goes from 0 to 1 by definition, so the correct calculation is:

(4*0.3-(1-0.3))/4 = 12.5% of the current balance

In other words, divide the probability you entered, 30, by 100. I only made the range 1-99 so that it would be more convenient to type.

so as to say if i apply it on roulette or other gambling platform or trading strategy, i juz need to apply those figures and i get the optimal trading size?
  • #7
  • Jul 25, 2009 2:52am Jul 25, 2009 2:52am
Here's also a Kelly Ratio simulator:

http://www.hquotes.com/tradehard/simulator.html
  • #8
  • Jul 25, 2009 2:58am Jul 25, 2009 2:58am
Quoting Holypendant
so as to say if i apply it on roulette or other gambling platform or trading strategy, i juz need to apply those figures and i get the optimal trading size?
you can use this calc to calc trading size. but it's not interactive.
http://www.forexfactory.com/showthread.php?t=180628
  • #9
  • Edited 4:04am Jul 25, 2009 3:53am | Edited 4:04am
Quoting Holypendant
so as to say if i apply it on roulette or other gambling platform or trading strategy, i juz need to apply those figures and i get the optimal trading size?
Yes. But in roulettes and all other casino games, expectancy is below zero, which means that, on average, you lose, so the system will ALWAYS tell you to bet zero (which is the right choice: casino managers are not stupid!)

In a trading strategy, this applies only if for every single trade you make (i. e. every setup) you have the same winning probability and R/R ratio, and you know them with great accuracy. There are also good reasons to bet LESS than Kelly. If you take a look at the maths, you'll see that on average Kelly performs well, but the variability of your results is huge, so you must be ready to withstand enormous drawdowns, well over 90%. Combine that with a bad estimation of your trading parameters, and you're toast.
  • #10
  • Last Post: Jul 26, 2009 2:36am Jul 26, 2009 2:36am
Quoting outspan
Yes. But in roulettes and all other casino games, expectancy is below zero, which means that, on average, you lose, so the system will ALWAYS tell you to bet zero (which is the right choice: casino managers are not stupid!)

In a trading strategy, this applies only if for every single trade you make (i. e. every setup) you have the same winning probability and R/R ratio, and you know them with great accuracy. There are also good reasons to bet LESS than Kelly. If you take a look at the maths, you'll see that on average Kelly performs well, but...
I seem to have forgotten the most important thing, which is: Kelly assumes there is NO serial dependence between consecutive bets, i.e., the result of the i+1-trade is not influenced by the i-th trade. This is NOT true with many trading systems!
0 traders viewing now