Calculating run-outs

[sq764]

So let's say I wanted to focus on a long term 'investment-type' return by finding an angle that allows 100 plays a month and is showing a 1.12 ROI over 5,000 races. Let's add in a 6% rebate to that ROI so we're at 1.18.

Let's also say this play is hitting at a 20% flat win rate.

With that being said, if I were to start with a $2500 bankroll and play $25 per play, what would be the best way to figure what my potential longest runouts would be and my potential for every tapping out completely?

[DanG]

Let's also say this play is hitting at a 20% flat win rate.

With that being said, if I were to start with a $2500 bankroll and play $25 per play, what would be the best way to figure what my potential longest runouts would be and my potential for every tapping out completely?
Far greater students of math then myself on this board, but just in terms of bet size to hit rate…

100 potential trials ($2,500 / $25) at a 20% hit rate has the theoretical run out of 21 gambles. Of course you would drastically alter this equation if you used “percentage of bankroll” instead of flat betting.

[sq764]

What type of sample size do you guys feel is reliable in trusting a particular ROI?

[DanG]

What type of sample size do you guys feel is reliable in trusting a particular ROI?
Again; I’m no statistician, but to me sample size “confidence” is a sliding scale with hit rate.

The ‘Hyper Favorites” in HTR at a 50% hit rate are going to be pretty darn near that level 20 years from now…where as a 8% bomb seeker will have huge volatility.

The higher the hit rate the less sample size is needed for confidence forward imo. 300 trials is what I use with a -10 point ROI buffer to allow for the inevitable hoard (us) :D that is also pouring over results.

The average fld size is 8 runners so the “random” average hit rate is 12%. I would try for 300 trials when your hit rate is 12% and adjust sample size down as you hit rate rises and vice a versa of course.

[DanG]

Far greater students of math then myself on this board, but just in terms of bet size to hit rate…

100 potential trials ($2,500 / $25) at a 20% hit rate has the theoretical run out of 21 gambles. Of course you would drastically alter this equation if you used “percentage of bankroll” instead of flat betting.

Of course that is "losing streak"...not run out; Sorry about that. :o

[sq764]

Of course that is "losing streak"...not run out; Sorry about that. :o


Yeah, maybe the better question would have been to ask what size bankroll I would need to avoid a runout making $20 wagers.

[km]

good discussion, thanks sq for your questions

I agree with Dan, spot plays are not like a roulette wheel where numerical expectation is constant in the long run

The LEARN MORE option in the Robot is a great device for understanding the highs and lows of your spot play. If you have a year's worth of data in the test, the individual calendar month's are displayed in the output. Spot plays with 20% or lower winners will have some major swings month-by-month, even zeros in some periods - and then boom!, it hits with a 2.75 ROI some month. But rarely is it going to be consistent.

Good rule of thumb, if the spot play doesn't show an ROI > 0.90 with at least 8 of the 12 months, then pain and suffering may occur before profits!

Also look at the Place ROI, it may be the saver in many months when the win is unlucky.

[sq764]

Ken, that is an interesting point about the place ROI.. I hadn't really thought about it in that perspective (lower place price vs. the actual place ROI)

[Bob]

Ken, totally agree about 2nd place. When looking at my own spot play candidates, one thing I look at is how the spot play does in the month (or months) that it is not profitable, how is it doing in the 2nd spot? If it is "knocking at the door", I will consider and study it further. That is, until you finish work on your "who wins headbobs at the wire" rating! :D