Arbusers

Hey guys, need some advice.

Here’s my overall stats for the last 4 years:

Stake: 83,125 units
Expected profit (vs. Pinnacle closers): 4,847 units
Actual profit: 4,995 units (6%)
Bets: 12,496
This year, I've experienced a major downturn, with a loss of 283 units over 2,000+ bets (-2.26%, while the expected profit was 5.9% against Pinnacle closers). Nothing significant changed in my approach, except I started betting more on underdogs and lower-tier competitions (Pinnacle limits under 1,000 euros). But even when I filter out underdogs and these lower-tier games, the profit doesn’t improve, so this slight shift in strategy doesn’t seem to explain the losses.

I've included a couple of graphs for reference:




I’m concerned about the trend. I knew my winning streak wouldn’t last forever, given that my bets had overperformed for months, even years. But for the drop to happen so suddenly was unexpected, leaving me doubting myself and even considering quitting. I've had 9 negative months over this period, with 5 of them this year (including the two worst months in my dataset).

I've analyzed this dataset extensively, but there’s no clear cause. It appears to be a mix of luck balancing out, the inherent volatility of long shots, not fully accounting for the favorite-long shot bias, and perhaps relying too heavily on Pinnacle prices in the lower markets. There’s some evidence for each factor, but it’s tough to separate cause from effect, as the downturn spans all areas (market size, outcome probability, etc.).

I have all Pinnacle closing lines and details for each bet during this period, so any suggestions for additional tests or comments on the data are more than welcome!

I love this kind of posts so I thank you massively for posting.

From the looks of it, it seems to me that something changed in the beginning of 2024. This is when your trend started deteriorating negatively. So you must search at this period and see what was made differently.

But at the same time, you are not giving a lot of information and I can only speculate on what went wrong. For example, it is not clear to me how do you determine value? Are you simply trying to bet as a higher price that Pinnacle's closing line? And where are you placing the bets, in Pinnacle or in softs? I could go on with questions for very long, so I am asking you to speak some more about the hidden aspects of your project.

Again, thank you!

Thank you for the kind words

For example, it is not clear to me how do you determine value? Are you simply trying to bet as a higher price that Pinnacle's closing line? And where are you placing the bets, in Pinnacle or in softs?

That's correct; I'm betting at higher prices on soft bookmakers, with all bets placed pre-match. I hope you understand that I can't disclose specific details on where exactly. Feel free to ask about anything else.

From the looks of it, it seems to me that something changed in the beginning of 2024.

I have the same intuition. However, there's no strong evidence that anything changed, including Pinnacle’s closing line efficiency. In my dataset, there's no statistically significant deviation from zero profit when flat betting on closers. I did find some limited evidence for a favorite-long shot bias on odds above 5.00 (not fully accounted for in my vig calculation at the time) and inefficiencies in markets under 1k, but nothing that would account for such a substantial profit drop.

I'm fully open to suggestions and would be more than happy to test any hypothesis you come up with.

congratulations on such good performance, depending on what your unit bet is, you clearly made a lot of money. As you certainly know, almost nobody does in sports betting. It appears you've been break-even since mid 2023 till now, with a large sample size of bets. Yet extremely profitable prior to that with an even larger sample size. I can see why you are confused. If nothing has changed, this money machine should be working, yet isnt. If you are making the same # of bets per month with the same requirements, and have flat-lined for 16 months - its hard to explain. If you have been making more bets than usual , it could be you became convinced of your success and subconsciously expanded your parameters, leading to a mix of profitable bets and negative EV bets.

The probability of the initial 8000 or so bets being consistently profitable seems way below .05.

Not sure if this will help in anyway with the information given, but I had something very similar happen to me back in 2017 where I was heavily reliant on Pinnacle showing me perceived value in low liquidity games and high odds. Sometimes the arb value I was getting was 7-12% at some point but then I only bet the soft side.

That year was the worst for me and I got my bankroll destroyed.

Short answer: Noise.

You need a real convincing hypothesis to come to a causal explanation here.

so is the Pinnacle closing line perceived as being 'the truth', I know the Odds Jam fellow in his early videos gave it god-like importance. Yet over on Reddit, a lot of people using Odds Jam say they have not profited using their info.

Can you filter results for time before match start? Im guessing you are bleeding in games that is more than 24 hours before match start where in my experience many Pinnacle lines are incorrect.

if you filter out underdogs and lower-tier games how many bets did you place this year and what is the average odds? Also how to you calculate EV, I mean what kind of formulate do you use?

James Ravazzolo wrote: ↑

Thu Oct 31, 2024 7:43 pm

congratulations on such good performance, depending on what your unit bet is, you clearly made a lot of money. As you certainly know, almost nobody does in sports betting. It appears you've been break-even since mid 2023 till now, with a large sample size of bets. Yet extremely profitable prior to that with an even larger sample size. I can see why you are confused. If nothing has changed, this money machine should be working, yet isnt. If you are making the same # of bets per month with the same requirements, and have flat-lined for 16 months - its hard to explain. If you have been making more bets than usual , it could be you became convinced of your success and subconsciously expanded your parameters, leading to a mix of profitable bets and negative EV bets.

The probability of the initial 8000 or so bets being consistently profitable seems way below .05.

James Ravazzolo wrote: ↑

Thu Oct 31, 2024 7:43 pm

Not sure if this will help in anyway with the information given, but I had something very similar happen to me back in 2017 where I was heavily reliant on Pinnacle showing me perceived value in low liquidity games and high odds.

Thanks.
The number of bets per month has actually decreased, but I don’t think that’s relevant for this analysis. We’re still talking about 2000+ bets during this drop, regardless of the exact timeframe. I did widen my parameters somewhat due to limited opportunities in lower odds ranges and high-limit competitions, but there’s no evidence that this has significantly impacted my results.



Just to note, the total number of these bets (Pinnacle limits below 5k and odds taken above 5.00) across the entire dataset is 379 (38, 72, 187, and 82 for 2021, 2022, 2023, and 2024 respectively), making it impossible to draw concrete conclusions. As for the initial 8000 being so profitable, I agree on the p-value but have no explanation for it.

schollionär wrote: Short answer: Noise.

You need a real convincing hypothesis to come to a causal explanation here.

Agreed. But it’s hard to ignore what the “eye test” is showing—that there’s definitely something going on. I'm close to ruling out all three initial hypotheses: 1) long-shot volatility 2) betting into EV- bets, and 3) bad luck in games with the highest EV+ and thus highest stakes, since I use the Kelly criterion (ruled out)

James Ravazzolo wrote: so is the Pinnacle closing line perceived as being 'the truth', I know the Odds Jam fellow in his early videos gave it god-like importance. Yet over on Reddit, a lot of people using Odds Jam say they have not profited using their info.

I don’t have experience with Odds Jam or similar services, this project is solo 100%. Efficiency can depend on market size and odds range. If you go to football-betting-data.uk, and run tests on their datasets, you’ll find that for reasonably sized markets and favorites to small underdogs, it’s accurate most of the time.

AnnaPrentice wrote: Can you filter results for time before match start? Im guessing you are bleeding in games that is more than 24 hours before match start where in my experience many Pinnacle lines are incorrect.

Thanks for the suggestion. Since day one, I’ve been filtering alerts by time before match start: 12h, 6h, and 3h for big, medium, and low markets, respectively, to better fit a recreational bettor’s style. While I haven’t tested Pinnacle line efficiency across time on a large dataset, I can say from my data that efficiency is nearly the same between the closer and the line available when I place a bet.

mrJustice wrote: ↑

Fri Nov 01, 2024 1:47 pm

if you filter out underdogs and lower-tier games how many bets did you place this year and what is the average odds? Also how to you calculate EV, I mean what kind of formulate do you use?

Good questions.
odds > 3.00 + limits < 5000 removed: bets=472, avgOdds = 2.13, profit= -240 units
I assume you’re referring to methods for removing the vig. When I first started, I used the simple additive method, but I later realized it wasn’t the correct way to do it. I then switched to the Shin method, and last year to the “proportional to odds” method, as recommended by Joseph Buchdahl: https://www.football-data.co.uk/The_Wis ... pdated.pdf

Filtered bets in your latest graph show there is room for you to play.

I wanted to say something about Pinnacle, that I mentioned before in another thread. Pinnacle closing lines are not divine. Over the years Pinnacle managed to build a reputation through clever marketing and articles by ''so called'' experts, most of whom have no real betting record, but they are based in mathematics and theory. I ve got news for them, the real world is different.

So here is my question: If indeed the Pinnacle closing line is divine, why don't they (Pinnacle) offer bigger limits? These last years we noticed Pinnacle pulling out of markets and lowering their limits. Pinnacle limits are a fraction of what they used to be 10 years ago.

I also noticed reports from members who were able to beat Pinnacle traders in tennis and other sports. I m sure Pinnacle's odds makers and traders are doing a terrific job, but I m more sure there are people in the other side of the fence who do an even better job.

Now coming back to the OPs story, I would examine the possibility that you are a victim of multiple value traps, this is something that we discussed extensively in the past. Maybe you could play a little bit with your statistics and see what are the bleeding points?

I recommend you to pay attention to all the market, not only Pinnacle. I mostly reject bets when Pinnacle is absolutely alone.

arbusers wrote: ↑

Fri Nov 01, 2024 4:15 pm

I wanted to say something about Pinnacle, that I mentioned before in another thread. Pinnacle closing lines are not divine. Over the years Pinnacle managed to build a reputation through clever marketing and articles by ''so called'' experts, most of whom have no real betting record, but they are based in mathematics and theory. I ve got news for them, the real world is different.

That is good advice. In my opinion the critical thing here is. A lot of data-minded people these days try to treat betting as it would be physics when in reality its more like psychology. Pinnacle did a good job here in establish a certain way of thinking about probability.

arbusers wrote: ↑

Fri Nov 01, 2024 4:15 pm

Filtered bets in your latest graph show there is room for you to play.

I wanted to say something about Pinnacle, that I mentioned before in another thread. Pinnacle closing lines are not divine. Over the years Pinnacle managed to build a reputation through clever marketing and articles by ''so called'' experts, most of whom have no real betting record, but they are based in mathematics and theory. I ve got news for them, the real world is different.

So here is my question: If indeed the Pinnacle closing line is divine, why don't they (Pinnacle) offer bigger limits? These last years we noticed Pinnacle pulling out of markets and lowering their limits. Pinnacle limits are a fraction of what they used to be 10 years ago.

I also noticed reports from members who were able to beat Pinnacle traders in tennis and other sports. I m sure Pinnacle's odds makers and traders are doing a terrific job, but I m more sure there are people in the other side of the fence who do an even better job.

Now coming back to the OPs story, I would examine the possibility that you are a victim of multiple value traps, this is something that we discussed extensively in the past. Maybe you could play a little bit with your statistics and see what are the bleeding points?

Good points. Their lines aren’t perfect, but on average, they’re closer to perfection than any other bookmaker. If the true no-vig odds are 1.75 and Pinnacle’s no-vig odds are 1.70 while a soft book has them at 1.85, it’s still a solid bet. This would typically be in a market like Finnish hockey rather than the NBA, NFL, or EPL, so we have to built in some safety margins in these smaller markets.

As for value traps, they've been around for a long time. Even if I knew how to spot them in backtesting, I don't think they've become more common in 2024.

Anyway, I’ve also ruled out the second hypothesis (betting on EV- bets) since I pinpointed the subset of bets responsible for all losses—and then some (not sure why it took me five days of analysis to figure this out). Interestingly, these bets are in markets above 5k.

Here's the breakdown (2024):

Markets < 5000: 1,453 bets, stake 8585, +282 units, expected profit 6.1% mean odds 2.66
Markets > 5000 (mean 15700, median 10000) : 579 bets, stake 3911, minus 568 units, expected profit 5.44% mean odds 2.62

These >5k markets mainly include the top 5 soccer leagues, plus tennis and American sports.

This tells me that the noise is the main contributor to the losses, because we all agree that these markets are super accurate on Pinnacle, right?
I am open to counterarguments and evidence to the contrary.

CharlieSheen99 wrote: I recommend you to pay attention to all the market, not only Pinnacle. I mostly reject bets when Pinnacle is absolutely alone.

Thanks for the advice.
I agree and I do check other books.
Long shots for instance are clearly more accurate (higher) on Betfair.