Spent last month going back through my records comparing edge quality across five-year periods.
2005-2010: average closing line value on my bets approximately 3.2%. Meaningful edge. Consistent.
2010-2015: 2.7%. Still positive. Noticeable decline.
2015-2020: 1.9%. Getting thin.
2020-2024: 1.4%.
The trend is clear and it's been clear for a while. The edges I was finding in 2007 don't exist in 2024. Not because I've gotten worse. Because the market has gotten better.
AI odds compilation. Official real-time data feeds. US institutional money improving European market efficiency. Syndicates operating at larger scale with better models.
The question I'm trying to answer honestly: is there still genuine positive expected value available for a serious retail bettor in football markets or are we past the point where that's realistic.
The public money edge I've built everything on has compressed.
Not disappeared. Compressed.
2010: fading heavily public-sided games produced strong returns. The gap between public perception and actual probability was large enough to be systematically exploitable.
2024: the gap still exists. The market corrects for it faster and more completely than it used to.
The books have gotten better at pricing public sentiment into their lines before I can act on it.
The opening line that used to sit at -3.5 public favorite in 2010 now opens at -4 because the book has already priced in the public lean.
The edge hasn't disappeared. It's been partially absorbed by the operators before I can reach it.
The window is shorter. The return per bet is smaller. The methodology still works.
For how much longer is the honest question I don't want to answer.
Specific data: my model's closing line value has decreased approximately 0.8% over the last seven years.
Still positive. The positive return is genuine.
But the trajectory is consistent with Eddie's observation.
The reason in my specific case: Bundesliga data infrastructure has improved dramatically. Stats Perform now provides granular in-play and pre-match data to operators that was unavailable seven years ago.
The information I was ahead of the operators on in 2017 the operators now have access to as standard.
My model compensates partly by going deeper into secondary factors the data providers don't capture.
But the data providers are expanding their coverage continuously.
Every year there are fewer factors they don't capture.
From the exchange the efficiency increase was visible in real time.
2012-2015: genuine pricing errors in specific markets. Observable. Exploitable. Consistent source of edge for disciplined traders.
2015-2019: errors becoming rarer. Corrected faster when they appeared. Window to act narrowing from minutes to seconds.
2019-2023: the obvious errors are largely gone from top-flight markets. The remaining inefficiencies require either significantly better data or significantly faster execution than retail bettors have access to.
The exchange premium charge tightened simultaneously with the efficiency increase.
The combination: fewer edges available, smaller margins when found, higher platform costs for consistent winners.
Left the exchange partly because the business case for serious trading was deteriorating structurally.
The US legalization is underappreciated as a factor in European market efficiency.
2018 onwards: serious American institutional money engaging with European football markets.
European football was relatively inefficient compared to US sports markets partly because the US capital wasn't in it.
Post-PASPA that changed.
Sharp American money is now consistently in Champions League and top Premier League markets.
Markets that were somewhat soft in 2017 have been bet to significantly tighter margins partly because of American capital and modeling sophistication entering them.
The domestic UK market tightened itself. The international market tightened because America arrived.
I don't find edges so this doesn't affect me the same way.
But I'm wondering about something.
If the serious analytical bettors are finding it harder to find genuine value, and they're the people who make the markets more efficient by exploiting errors...
Does the efficiency increase actually slow down at some point because the edge-finders are less profitable and therefore less active?
The people creating efficiency are being squeezed out.
The destination question is worth examining directly.
If AI compilation continues to improve and data coverage continues to expand: at what point is there no edge available to a retail bettor regardless of methodology?
My honest assessment: top-flight European leagues may reach that point within five to seven years for standard markets.
Lower leagues, specific market types, and markets with genuine information asymmetry will retain edges longer.
But the trend is not reversible.
The retail bettor serious about positive expected value will increasingly require either specialization in thin markets or access to tools that approach institutional quality.
Klaus's five to seven year estimate for top-flight market efficiency.
I think that's optimistic for the top and pessimistic for the edges.
Champions League and top six Premier League? Already close to uninvestible for retail.
Championship, League One, lower Bundesliga, second-tier European leagues? Still meaningful inefficiency.
The shrinkflation is happening from the top down.
The strategy becomes progressive retreat into thinner markets where the data infrastructure is less developed.
You follow the inefficiency downward until you hit the floor.
The floor exists. It's probably the level where the official data providers don't cover and the liquidity is too thin for institutional money to operate efficiently.
Thirty years of watching this happen in real time produces a specific perspective.
The edges I found in the early nineties were different in kind not just degree from what's available now.
Genuine pricing errors that persisted for days in some markets. Information advantages from actually knowing people close to clubs. Price discrepancies between different operators that no one had modeled.
Those edges are gone. Have been gone for fifteen to twenty years.
The edges I found in 2005-2010 were different from those. More sophisticated. Required better methodology.
Those edges are substantially gone now.
What remains requires better methodology again.
The pattern repeats. Each cycle of edge requires more sophistication to find and less return when found.
I have kept up with each iteration partly through discipline and partly through thirty years of accumulated pattern recognition.
I don't know how many more iterations I have in me.
At fifty-eight the question of whether the next level of sophistication required is within my reach is one I haven't answered honestly.
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.
