You might want to increase "Time to match start"

We recently increased the time to match start default value from 12 hours to 48 hours. This only affects new users.

The reason is we found no correlation between time to match start and (flat) profit.

There was no correlation between: closing line value, flat profit or wether you beat the closing line. We checked 330.000 bets with closing line.

This was a bit surprising, since the “common knowledge” is that value bets get more reliable closer to match start (all things being equal). Of course, all things are rarely equal :slight_smile: Anyone have a hypothesis?

The new recommendation is: increase time to match start as high as you can while still maximizing turning per day. You don’t want to let your bank roll sit unused while you wait for matches to end.

This might be 12 hours for users with smaller bankrolls, or 300 (our maximum) for those with large bankrolls.

1 Like

Hi Simon,

This is a really interesting result. Thank you.

  1. When you run the same analysis for bets from an individual bookmaker (for example, Bet365) do you arrive at the same result? I wonder if there is any finer structure when doing the analysis by bookmaker instead of using the data set with 330000 bets (which I guess includes bets from many bookmakers).

  2. When increasing the time to match start value, would you expect the variance to increase?

Many thanks,

AXB

  1. Yes, very similar numbers with Bet365.

  2. That’s an interesting idea, sounds likely. I don’t have a quick way of checking right now, have to save that one for later.

Betting closer to match start increases the chance of beating the closing line very slightly. Which makes sense. But the correlation is just 0.10, too small to care about.

2 Likes

Is it a good idea to sort time soonest before kick off to show on top, while keeping 48h to start filter too?

That way we prioritise the soonest to start ones giving us a chance for return of money in balance quicker?

I believe it makes sense. I’m thinking about using something like value/(time before bet + match duration) to keep the turnover high. It’s better to bet 3 times on 3% value bets happening soon than to bet once on a 8% value bet that’s happening in 40hours.

A smarter sorting would be nice, but it might be difficult to figure out exactly how it should be calculated and presented.
If you have two value bets, one 3% starting now and one 6% starting in one hour, which one should be sorted at the top? It depends on how many value bets you usually get each hour, so it quickly gets complicated. And you also optimally want to bet on both of them!

Easiest right now is to have filters, one pinned to the top with matches close to match start.

Let’s say the total time for a match to be played + its grading is 2 hours, the value per hour would be:

3% / (2+0) = 1.5% for the first match starting now (time = 0)
6% / (2+1) = 2% for the second match starting in one hour (time = 1)

Therefore the second match would be at the top.
This technique would however require someone to be betting more frequently to make full use of it and, like you said, enough value bets.

Thank you for analysis like this.
Does this analysis consider the projected value at the time the bet was made? “Flat profit” makes it sound like the absolute profit of betting 500 units at 200hours returns the same units as betting 500 units at 15 minutes, regardless of the value when the bet was made.
However, is there a trend between the delta of the placed EV vs the CLV and the time to match.

In summary, what percent of the projected value do I realize (vs CLV) based on time to match.

There was no correlation between time to match start and value at the time of bet placement.

Flat profit might be a confusing term. I mean calculated with a flat staking plan, betting 1 unit on every bet (this removes the effect of staking strategy).

You’re right that you should always bet closer to match start if you have to choose. If you don’t turn over your entire bankroll each day, you should increase the time to match start filter.

Interesting last two questions, I’ll have to take a look!