KC vs Flatbetting

When you have an edge, 5% for example you must decide how will you manage your money.

Sometimes, when your numbers have won you could be losing if you use Kelly criterion because you raise your bet when winning and lower then on a bad stake.

The good part is tha when you get hits you win more.

Flatbetting would ensure your 5% edge but does not save your BR in case of a long dry.

Another items comes with the idea of dividing you BR in 2 3 or 4 parts and apply KC for the part you invest that day.

Is it the same to have 1% edge than 5% or 10% as regards as choosing flatbetting or KC?

There are more things to discuss about.

I’m not a mathematician, but my impression is that the only good reason to NOT apply the Kelly Criterion, is if you have doubts about your edge. I mean, how can you KNOW that you have a 5% edge? Is it 5% all the time? Isn’t it rather 6% or 4%?

And even then I’d advice you to strictly follow the mathematically correct adjustments to the KC, according to the change in its premises which you observe.

It hasn’t anything to do with assuming that good or bad streaks will continue, or revert. Changing bet sizes with KC only has to do with the fact that the size of you bank roll changes, and the betting size is a function of the size of your bank roll (correct me if I’m wrong).

Flatbetting would make you much worse off than using KC.

Dividing the bank roll in different parts might feel good psychologically. Like a squirrel feels good about digging down nuts in different places. But your instinctive stone age feelings don’t influence the suitability of your betting strategy, so you should ignore them and trust the math.

You have a accurate edge when you have tracked more than 15k spins but before that you guess your edge.

When you have 9% edge in 5k spins you estimate 6 to 7% just in case.

When you have over 15k decitions you know your edge.

I often use KC but I see that the days that numbers do not come as expected or more flatbetting is better. I do not say flatbetting is better but isn’t good not to win or lose when your numbers have won.

What I do is to divide my 10k BR in 4 of 2500 and apply KC with 2500 for the session.

Next session I take 3000 if the kast one was succesfull.

We know math does not lie.

This is my way but it can be improved, that’s why I opened this topic.

Best regards.

If you make a small sample test of negative progression (Martingale f.ex.) versus a positive progression like the Guetting progression, you will see that short term the negative progression will have a higher succes rate than the positive progression. Long term they are of course equal.

But its the same thing you face when you apply the KC, there will in the beginning be streaks where the KC seems to be the wrong choice. But you have to remember that the KC is considering 2 things:

  1. Keeping the bet size at a size so that the possibility of going bankrupt in the case of a negative deviation, despite an edge, is kept at a minimum.

  2. At the same time keep a maximum growth in the bankroll.

If you don`t need number 1, and just want the maximum growth, you might as well apply a negative progression. Which if you have an edge, actually is “okay”. BJ players swear to KC because with only +1% edge, the negative deviations are quite scary. Its a little bit different in roulette, because usually we target higher edge percentages and we can also tell if our edge is still there or not. Still, if we hit a casino with a 20.000$ bankroll, we still face a bankroll of 200.000.000 or something. Any mathematician will say: “Run, god dammit, run” even if we maintain that we have an edge of 10% over the house. So generally, as a safety matter, i would recommend KC because it considers both safety and maximum growth.

I would also recommend learning some sort of VB or light VB, not as a direct prediction but to filter out the spins you are pretty certain comes nowhere near your biased numbers. You could as a safety, just play with minimum bets on those spins, in case the unthinkable happened, then you will get some anyway.

If you consider dividing your bank in 4, you could try and backtest your number stream with a negative progression inside each 1/4 of the bank. If you bust out, you have lost 1/4 of the bank, if you win, you can either add it all to the 1/4 you are playing with or divide the profits in 4 and add size to all 4 parts.

But first backtest, then fine tune, and then play.

Generally, the higher edge you have, the better growth will the KC produce. On the contrary, if you have rock solid cash flow when flat betting, you might wanna consider a negative progression with a very flat rise in bet size. Just don`t throw rocks after me if you go bankrupt after busting out with high stakes after a negative progression, because i would probably say: “I told you so”.

I’ll keep on using KC.

No chance for VB, possibly dealer signature as the rotor always goes CW and the ball goes ACW.

I’ve been studying a possible signature, the trend seems to go ACW from the last number spun.

The test was done in about 2k spins(not many) taking the last spun number and 17 numbers ACW.

The idea is to weigh the numbers that are in the sector 17 numbers ACW and lower the others.

Another scenario:

11 numbers with 5% edge. All of them are neighbors.

The 11 numbers do not have the same edge one by one.

For instance, 4 of them have 9% edge, 4 of them have 4.5% and 3 have 0.50%.

Applying KC to all of the it comes to:

(35-10)/11
A=25/11=2.27
F=5%/2.7=1.85% of my current BR divided on the 11 numbers.
BR10K= 17 each

What happens if we apply KC to the 3 edges separately?

We divide our BR too, 3300x3=9900(it could be divided in other way)

4 numbers with 9%: 9%/8= 1.125% on 4 numbers= 9 units each

4 numbers with 4.5%:4.5/8=0.56% on 4 numbers= 5 units each

3 numbers with 0.50%: 0.50/11=0.045 on 3 numbers= 0.50 each

You take every edge with different money management and raise or lower your bets depending on the set of 4 of 3 numbers that hit or miss.

I suppose the theoretical correct way to do it would be as you describe it. And if you can get a modus operandi for the betting sizes put down into an easy to apply scheme you might get away with it. My first thought though, was that it would be hard to put in action in a practical manner. But it does make sense to have some sort of differential betting since there are so much difference on the edges from number to number.

That’s why I thought it.