Is Kelly Staking Perfection?

Is True Odds Flaw The Only Worry?

Nope.

You the punter are not a machine either.

There is an old joke:

"The Police report stated that the car accident was the result of mechanical failure. The nut behind the wheel was at fault."

The same is often true of a punter with a good method and a decent staking approach. The weak link in the chain is often the emotional human at the controls.

Kelly staking using the full Kelly suggested percentages will produce some significant up and down swings in one's betting bank. Kelly maths is aimed at optimizing long term growth potential and not worried about short term volatility or the impact of said volatility on the human mind.

Some nut might fail on the way up where an excess purple patch shoots their balance up quickly. Temporary delusionary "I am clearly a God" syndrome can lead to some silly life choices. Perhaps in hindsight it was not such a great idea to tell your boss that he both looked and smelled like a pig. Highlighting to him which of his orifices the job should be shoved, upon reflection you now consider a tad rash and premature.

More humans however are likely to fail during a poor patch, most especially if said poor patch comes near the start of their venture.

The higher the aggression level of staking one sets, the more violent the up and down swings will be.

So for long term success it is probably quite important to understand your own unique self.

Some individuals may have extreme emotional detachment from their betting bank. They won't care about days or months of ups or downs and will instead have laser focus on a few years down the road. Others will sweat every day or every poor patch. Each day is a tortured emotional experience. A couple of losers in a row and their brain will turn into chicken licken believing the sky is falling down.

Statistical scatter is a real thing. The volatility it causes in one's bank can however be controlled to a degree by the level of aggression one takes.

Fractional Kelly stakes is one common concept. One can use say 50% of what Kelly suggests. As discussed above that will bring in a degree of safety factor as some error in true odds calculation must be assumed. It also helps smooth the ride a little in terms of how violently ones bank rises and falls over time. With less violent up and down shaking, the nut behind the wheel is less likely to fail.

Even if your Kelly stake is fractional, you will still have the bank management benefit of staking ratios. Higher edge & higher strike rate opportunities merit increased aggression over lower strike rate / lower edge options.

So whilst it is good to gain a feel for the pure maths of it, arguably more key in the long run is to understand your own unique emotional make-up and the degree of volatility you may be able to work with.

Can Kelly Help Grow My Horse Racing Bank To Infinity?

Well it can certainly assist in growing it but infinity may be one step too far to hope for.

One can analyse betting data in a spreadsheet and marvel at how the right compounding staking approach would have grown £1000 to £2k. Then to £4k. Then to £8k. A few jumps later it may be at £1 million. But only in the land of the spreadsheet however.

Bookmakers tend not to be super pleased to keep accepting bigger and bigger stakes from those clients who keep winning with them. So early doors in one's bank building having an eye on Kelly style principles can be a good thing to get your bank built up from small to much bigger at a semi-efficient rate.

However, some future day one may have the problem of the stakes you put down not being set by what your own calculations dictate. Instead, they may be lower than desired and capped by the real world constraint of what you can actually get down at an acceptable to you price.

That is just the pragmatic realities of the real world. On the scale of world problems, it is not the worst. To arrive in the problem position of not being able to bet your full Kelly stake, you have probably done OK for yourself on the path to get there.

What Problems Does Kelly Have With Horse Racing?

Kelly Formula was designed to work with simple binary style bets. ie Bets with a simple yes or no style of outcome. So it is perfectly apt for simple bet the horse to win style occasions. Did Dobbin win the race? Yes or No will be the answer. [please bog off Mr Pedantic who wants to chirp up about dead heats]

Kelly handles that great. The same scenario with a betting exchange place only market. 1 or 0 The horse placed or it did not.

A win double or treble? Again no problem. It wins or it loses.

A common horse racing bet not super apt for Kelly staking is our old friend the each way bet. Here we have three outcomes: The Win, the loss, and the place. What are the bookmaker odds for our Kelly equation here? Well, it is a square peg in a round hole style scenario. There are two payout scenarios for the horse wins and the horse places scenarios. The place scenario as well may not even produce a net profit. 9/2 placed at 1/5th odds, for example, is a very small loss.

The Kelly Formula will just whirr around in confusion as such concepts it was never designed to accommodate.

Another common enough racing scenario is the concept of betting more than one horse in a race. This is a concept that can increase strike and reduce the frequency with which you lose your whole stake. It can have a fair amount of logic going for it from a bank building perspective if you add in the idea that reduction in volatility can permit more aggression and thus a win adds more to the pot. But again due to having multiple potential payout scenarios (unless you happen to perfect dutch), it is an imperfect match to fit into the slots of the Kelly equation.

So if you have each way bets as well as win bets or sometimes like to cover two in a race what do you do? Do you take your profitable method and mess it all up just to try and get it to work with a maths equation?

Nope I say to that idea. Instead, forget pure Kelly. Aim instead for something in the Kellyesque zone of things. Perfection is not required. You are going to be slapping a big fudged safety factor on it at the end anyhow.

What I like to do is start with a historic dataset in a spreadsheet. It could be a list of your own past bets, some researched methodology or system you have defined. Perhaps even it is a list of historic tips from your favourite tipster.

Set up your spreadsheet to consider compounding a bank using a fixed % stake. One can then alter that % box in the spreadsheet to see where the end bank ended up. What percentage X as stake produced the biggest bank growth?

You will find that for low strike rate methods the X will be small. It will tend to be higher for higher strike rate / lower bank drawdown approaches. Similar with the issue of edge at level stakes.

So it is some what Kellyesque in approach. However instead of pondering an individual bet it is more so pondering a longer duration dataset that has odds, strikes, drawdowns and edge.

When you do get your end result magical figure of X% that produces the most wondrous bank growth that was possible, what do you do then?

Well just as per Kelly I would apply safety factors to it. For pretty much all the same reasons. This time your true edge is more crystal clear as it is based on the cold hard numbers of history. But it is based on history, NOT on the future, so sandbag it a bit and build in a safety buffer to move forwards with.

The human emotional side of things is pretty much identical. A smoother less volatile ride helps most humans.

The above idea is simple enough. It will give you a % that you will use across all future bets of the same method. Usage then is relatively simple as you would always be working to the same % stake.

Could that be refined further? Yes. It is not an idea I have used myself as yet but one could treat the given percentage as a base percentage. You then move it up or down a bit for individual bets depending on the precise odds of the individual bets.

Some people will prefer simplicity, others will seek higher levels of complication.

Key Point Summary

• Kelly Staking is a useful mathematical formula to determine the theoretical optimal percentage of one's bankroll to stake on a bet.
• High edge and high strike rate lead to higher aggression levels than low edge and low strike rate.
• Kelly will advise you to bet £0 if there is no positive edge in your favour.
• Over aggressive staking does more harm than under aggressive staking.
• Add safety factor (or use fractional Kelly) to account for flaws in your true odds assessments.
• Be aware that the weak link in the chain is probably your own human emotion. Design for that by adding safety factor or fractional Kelly to smooth out bank volatility.
• Kelly only works for simple win or lose style bets.
• For horse racing betting that also includes each way or multiple horses covered, forget about pure Kelly and instead aim for a Kellyesque approach.

Another article that uses such a Kellyesque approach to examine real life racing betting data may be found here Win Or Each Way

Mick