Kelly criterion when having pending bet.

[Dujkan]

Lets say I place kelly bet that took 20% of my bankroll (was 100, now is 80). And now I see new value bet, with odds 2.1 (50%chance of happening). How much should I bet from my remaining bankroll (80 at the moment), do I just use kelly criterion from 80 bankroll, or should I take into calculation that I have pending bet? and how?

[eaglesfan371]

Kelly would suggest not having more than 2-5% at risk for any wager so you’re already going against it.

[danshan11]

how do you know you have value?

the line is currently +100 and what do you think the line should be? thats one way
or you have a long term win% using the same method that wins at 57%? thats another way

[Dujkan]

Kelly would suggest not having more than 2-5% at risk for any wager so you’re already going against it.

If there are bets with odds 1.01 or 1.02 then kelly could suggest you to bet even more than 20%. And I gave that as example. You did not answer my question, when bankroll is down to 80, do you take that into calculation, or you just use kelly on 80 that is left. It must not be the same when you calculate kelly when you aleready have pending bet and you when you dont.

[Dujkan]

how do you know you have value?

the line is currently +100 and what do you think the line should be? thats one way
or you have a long term win% using the same method that wins at 57%? thats another way

You too didnt answer question. How I find that bet has value is not in the question here. When I lose a bet that cost me 20 of 100 bankroll, I just calculate kelly on 80 that is left. But how to calculate kelly when you have pending bet, I have 80 bankroll and a bet that has some percentage possibility of wining. How you take into consideration kelly then, how much you stake second bet with 80 points that are left from bankrol? (you could also say 80% of bankroll) It is hard to believe that you just calculate same kelly on 80 points bankroll when you have, and dont have pending bet.

[Optional]

I "think" if the bet is live it would as tough to estimate an accurate value as making live betting lines.

But there are a couple of good threads about calculating simultaneous bets with kelly that might help you with math ideas for it.

https://www.sportsbookreview.com/forum/handicapper-think-tank/521569-simple-closed-form-solution-unconstrained-simultaneous-bet-kelly-staking.html

[danwinkler]

Lets say I place kelly bet that took 20% of my bankroll (was 100, now is 80). And now I see new value bet, with odds 2.1 (50%chance of happening). How much should I bet from my remaining bankroll (80 at the moment), do I just use kelly criterion from 80 bankroll, or should I take into calculation that I have pending bet? and how?

Try this variation

[HeeeHAWWWW]

You do take into account pending bets. Assuming you're talking X independent events at the same edge, an approximation is that each should get reduced by (1-KellyStake)^(X-1).

If you haven't done this in advance, , you'll need to work through the whole tedious calculations each time.

Ditto if you have drastically varying edges

(a range is fine, but if one is 20% and the rest 1%........ nope).

Personally I just estimate how many bets I'll have running in advance - as long as you're in the ballpark the error makes almost no difference.

[Larkman]

If you aware of both sets of probabilities and odds at the time of betting, you need the simultaneous kelly criteron. If the odds for the second bet are not available until after placing the first bet on I'm not sure there is a mathematically rigorous answer.

In ordinary sequential kelly, you'd wait until the outcome of the first bet and then use the updated bankroll, so given that, using a % of your expected bankroll based on the expected return of the first bet seems logical but I'm pulling that out of my arse.

[Sam Odom]

jjgold wrote a White Paper on this very subject... Very in-depth and scholarly

Contact him

[RudyRuetigger]

You do take into account pending bets. Assuming you're talking X independent events at the same edge, an approximation is that each should get reduced by (1-KellyStake)^(X-1).

If you haven't done this in advance, , you'll need to work through the whole tedious calculations each time.

Ditto if you have drastically varying edges

(a range is fine, but if one is 20% and the rest 1%........ nope).

Personally I just estimate how many bets I'll have running in advance - as long as you're in the ballpark the error makes almost no difference.

[danshan11]

You too didnt answer question. How I find that bet has value is not in the question here. When I lose a bet that cost me 20 of 100 bankroll, I just calculate kelly on 80 that is left. But how to calculate kelly when you have pending bet, I have 80 bankroll and a bet that has some percentage possibility of wining. How you take into consideration kelly then, how much you stake second bet with 80 points that are left from bankrol? (you could also say 80% of bankroll) It is hard to believe that you just calculate same kelly on 80 points bankroll when you have, and dont have pending bet.

alfa did a good write up on this in the think tank recently. basically what he said is, you adjust your bankroll periodically at a set number of bets or a period based on the current bankroll

say you bet 1% (just for conversation not a real#) and br is 100, so you bet 1 dollar each on 10 games
now you lose 6 bets and your bankroll is now 8, you would not adjust down to 80 cents here, you would stay at a dollar until you have ran a cycle. once cycle is completed you now adjust your 1% to the new bankroll at that time. you dont want to adjust your bankroll based oin each result or what is currently in your bankroll.
I think alfa said 30 days or 50 bets was what he considered a cycle.

if you adjust your bet size too much or too often, your ROI can get messed up in a good or bad way, if you actually had an edge by variance

bankroll 1000
bad example 10 bets of 100
lose 7 bets now you have 600
if you bet the next 10 games and adjust to 60
you win 7 and now you have
7 losses at 100
7 wins at 60

you can see how in this scenario you went 50-50 assuming no juice and you are still down!
obvious what sucks about that and can work the opposite way as well

[TommieGunshot]

You do take into account pending bets. Assuming you're talking X independent events at the same edge, an approximation is that each should get reduced by (1-KellyStake)^(X-1).

Good stuff. Would like to see a formula that does take into account various edges. The variance of the edge must also be accounted for. Getting something at +1000000 with a 0.10% chance of happening is far different that something at -10000 with a 99.99% chance of hitting

In general, since most people bet fractional Kelly (and bets that generally have a realistic chance of winning or losing), not accounting for pending bets is almost always fine. In the above example using a bankroll of 100 is usually fine

[TheMoneyShot]

jjgold wrote a White Paper on this very subject... Very in-depth and scholarly

Contact him

Landers also wrote an in-depth summary on this subject....

He claimed "It isn't juice unless you lose."

Top notch

[HeeeHAWWWW]

In general, since most people bet fractional Kelly (and bets that generally have a realistic chance of winning or losing), not accounting for pending bets is almost always fine. In the above example using a bankroll of 100 is usually fine

It's only really a factor once you get into hefty volume betting. I had 50 simultaneous bets last week, so had to reduce stakes by approx 3x. That then gives you the same EG and volatility as a single independent Kelly bet.

For say 5 bets at once, it really doesn't matter much.

[u21c3f6]

It's only really a factor once you get into hefty volume betting. I had 50 simultaneous bets last week, so had to reduce stakes by approx 3x. That then gives you the same EG and volatility as a single independent Kelly bet.

For say 5 bets at once, it really doesn't matter much.

Holy cow HH, I thought I was nuts with the number of simultaneous wagers I make at times but 50? Congrats and wishing you good fortune. Joe.

[Miz]

i use a table to look up estimated kelly bet percentage based on estimated number of simultaneous wagers i expect... much like other folks here.

[TommieGunshot]

It's only really a factor once you get into hefty volume betting. I had 50 simultaneous bets last week, so had to reduce stakes by approx 3x. That then gives you the same EG and volatility as a single independent Kelly bet.

What odds? At around 50% with a 1% edge, Kelly says bet around 1% of your bankroll. Betting half Kelly, that means there's only 25% of your bankroll in action. Betting half kelly, with ~50% chance of winning each bet it would take three standard deviations to the bad side to lose 5% of your bankroll (it's even more likely you'll be three SD to the good side and be up 5%). With a big edge and betting a lot more I could see having to reduce bets, but at that point most everyone is making limit bets and still betting under full Kelly

[MC PICKS]

"I dont know who she is"
-Kelly In Vegas.

[thebaibol]

hahaha

[sweethook]

as you get older youl learn that in sports betting 99.99% of all events are situational . meaning if a thug was out thugin all night cuttin grass or makin it rain it will show up in the score . so if you can do the math on that sir. youl never have to work again .. gl

[Believe_EMT]

what you should learn as you get older is the humility to understand the market is chalk full of people much smarter than you, with deeper pockets than you and more/better technology than you. you're bet does nothing to the line, whether broke arse kelly says 1% or 200% of your BR. but those smarter and richer folks do.

let them do the work. grab your biscuits and sop up the their leftover gravy.

[hubie69]

jjgold wrote a White Paper on this very subject... Very in-depth and scholarly

Contact him

+350 if JJ can spell white paper.