Samuel's Blog

Adventures In Betting: Further revision to the greyhound selection criteria

February 29th, 2024 at 02:02pm

It would be fair to say that last week’s revision to the selection criteria for the greyhounds in the LayPro88 staking system didn’t quite go according to plan. The third favourites improved quite a bit and were roughly breaking even throughout the week, ending very slightly down. Good news but still room for improvement. The fourth favourites though started off badly (as seen in my day trip to Goulburn) and stayed there, sitting roughly level most days but with a downward trend overall.

Breaking it down into odds ranges tells a bit of a story

(HTML table formatting is a pain, it’s much easier to just take a screenshot out of Excel rather than paste the data in and fix up the HTML formatting, especially on a table this size).

It is notable that the strike rate on the higher odds for the fourth favourites dropped off by about 5-10 percentage points. This explains most of the trouble. With five bets in the recovery cycle following a loss, it is not possible to fully recoup a loss on anything paying over $8 without more winning bets after the recovery cycle. As such, the higher the odds, the higher the required strike rate, but at the same time the higher odds tend to improve the strike rate and the systems don’t turn a profit without something in the higher odds ranges. The dropoff in strike rate at the top end of the odds was costly. There will always be some variability in strike rate in any form of racing and it’s possible this was just a bad week, but it’s important to be able to handle a bad week and the performance in this week was worse than I would like to see from a bad week. Eliminating losses entirely isn’t possible, however reducing the damage they do is the aim so that the profitable days can far outweigh the bad ones should be achievable.

Based on this data, I have reduced the top end of the odds range from $15 to $13 for both the third and fourth favourites. The fourth favourites have a decent chunk of activity in that range, and losses are very costly up there. When the strike rate drops at that end, it is virtually impossible for the rest of the system to compensate.

I have made the same change on the third favourites. While the losses at that end of the third favourites remain rare, they are still costly. If things go well I might reconsider allowing the third favourites up to $14, but I’m not sure that the greatly increased strike rate up there is worth the risk of losses.

On the low end of the scale, the third favourites remain at a $6 odds minimum. This end continues to perform well despite a lower than desired strike rate as a five bet recovery cycle recoups more than the lost amount. The strike rate for the fourth favourites below $7 was poor when it was in use previously, and the handful of bets which got placed at the lower odds due to market flucturations at bet placement time remained at about the same strike rate, so the $7 minimum seems to be fine there.

So, starting today the new odds ranges have been set as:
3rd favourites: $6-$13
4th favourites: $7-$13

And another week begins…

Incidentally, I’m running LayGreyBot in a virtual machine along with a few other bots. While the bots themselves don’t use much memory, their memory usage does increase over time but tends to level off after a few days. Windows 10, even with as much of the unnecessary preinstalled bloatware disabled as possible, tends to take up a bit of memory and doesn’t always handle that expansion of memory usage from the bots very well. The host machine for the virtual machine is used for other purposes intermittently and doesn’t have a huge amount of RAM to begin with, so I originally only assigned 2GB of RAM to the Windows VM, which was fine with one bot when I started out, but with multiple bots running it started to cause unexpected application crashes and the odd full Windows reboot. 3GB was better but not great, I’ve found 4GB is fine for running multiple bots. My advice for Windows 10 is allocate 1GB for Windows and 500MB per bot, then allocate an extra 1GB if running more than one bot. I haven’t tested Windows 11 but wouldn’t expect it to be any more efficient.

At some stage I’ll probably look at moving the bots onto another machine with more resources. Whether that’s the existing VM moved on to more generous hardware, or whether I build a new environment as either a physical standalone machine or a new VM, I haven’t decided yet.

Samuel

Entry Filed under: Adventures In Betting

 Print This Post