Beating the Closing Line Doesn't Always Mean You're Good: Where CLV Misleads

ParlayPrincess_88

ParlayPrincess_88

Market Sharp
Joined
Aug 6, 2024
Messages
679
Reaction score
5
Points
8
Okay so I've been doing the thing where I write the line when I bet AND at kickoff (it was in the last thread lol). 41 bets so far. My average closing line value is +1.4 points. Everyone says that's the most important number!!
But I'm down 6.2% on those same 41. So am I good or am I just losing? 😭 And is CLV a real thing or one of those things people say on the internet?
 
Everygame
CLV is a real thing, but 41 bets can't settle it either way, for the good number or the bad one.

Your -6.2% first. At odds around 2.0, the 95% range on 41 flat bets is roughly plus or minus 30% from luck alone. So being down 6.2% says almost nothing yet.

Your +1.4 is the more useful number, but it is noisy too. For scale, the value flags we log have 1,546 settled and average +1.6% against the close, with a 95% interval of +1.1% to +2.1%. It took that many for the interval to clear zero clearly, so keep logging.

Two places it can mislead you. First, compare against the close with the margin taken out, because a raw comparison can flatter you. Second, use a price you could really have taken, not the first one you saw. In our log the +1.6% is measured at the moment we spotted the gap. For the 83% of flags we could re-check within 60 minutes, it drops to +1.0%, so over a third is gone by the time someone could act.

Disclosure: I run TipsAudit, which logs this publicly, so weigh that accordingly.
 
Princess, I will answer in two parts, because your question contains two numbers and the answer differs for each.
First, the ROI. At 41 bets the standard error on ROI is about 15 points. A result of minus 6.2% is statistically indistinguishable from plus 9% and from minus 21%. It says almost nothing. That is why closing line value exists: it is a far less noisy measure of price quality than results, and over a few hundred bets it predicts long-run ROI better than the ROI does.
Second, the CLV, which is where I need to be careful, because I am about to argue against my own favorite statistic. CLV is informative only when three conditions hold. The closing line must come from a market sharp and liquid enough to be close to the true price. The closing price must have the margin removed. And the sample must be large enough that a positive average is not luck. Violate any one of them and you get a number that looks like skill and is not.
The second condition deserves an example. You take Team A at minus 3, minus 105. The line closes at minus 3, minus 110 on both sides. You beat the close by five cents. The no-vig price at the close is 50%. Your bet pays 100 on a 105 stake at a true 50%, so your expected value is minus 2.4%. A positive CLV and a negative bet.
Do the math before you report the number.
Trust the process, not your gut.
 
Convert each side's price to a probability. Divide each by the total of both. What remains is the price without the book's margin.
 
Here's what nobody wants to hear: for a whole season my CLV was great and it was fake. I tracked it against the closing price at the book I bet most. Average +2.1 points across 214 bets. I put a chart in my sheet. I felt like I'd cracked it. (There was a chart. Of course there was a chart.)
Then I checked the same bets against the closing line at the sharper book, the one the soft books copy from. +0.2. Basically nothing. The book I was measuring against posts college football totals that lag by several minutes, so every time I bet a move I'd "beaten" a number that was just late.
I wasn't beating the market. I was beating the slowest clock in the room. RIP public money, but also RIP a very nice chart. Vegas knew a few minutes before my book did.
 
Welsh rugby on Betfair.
Midweek friendly.
Took 11/4.
Closed 5/2.
Spreadsheet says I beat the close, proper pleased.
Went and looked at the matched money after.
Whole market was three hundred quid. One bloke.
Beat one bloke butt.
That's not a close mun. That's a man having a tea break.
 
A close built from £300 of matched money is not a price. It is one person's opinion. I do not calculate CLV below £5,000 matched, which excludes about a third of my markets.
 
There is a bias in the close itself, and it matters for which bets one chooses.
The favourite-longshot bias is documented across decades of betting markets: the close slightly overprices longshots and slightly underprices favourites. A bettor who backs longshots will often beat the close and still lose, because the close he is beating is itself too generous to them.
In my own data, selections priced above 4.00 when taken: 61 bets, average CLV +3.4%, ROI minus 2.9%. Selections shorter than 2.50: 224 bets, CLV +1.9%, ROI +6.4%. The longshot sample is too small to prove anything. It is also exactly the shape the literature predicts.
The close is the best available estimate of the true price. It is not the true price.
 
As a coach, the best CLV I ever had was on a bet where I knew nothing. I took the road team getting 3 on a Wednesday. Friday the home team lost its starting quarterback, and the line closed with my team laying 1.5. Four and a half points of closing line value, the best number in my whole log. I'd picked them because I liked their run defense. I didn't know the injury was coming.
My sheet says that's the best bet I ever made. It isn't. It's the luckiest. News arrives in both directions, so over hundreds of bets it washes out. On one bet it tells you almost nothing.
I think there's a middle ground: CLV is a grade for the portfolio. I'd never put it on a single bet, and I stopped telling my wife about the big ones for the same reason.
 
lads the one number that has never flattered me is clv... mines negative every month i bother to work it out... and its not luck... its that im late... im always betting the price after the price everyone else already took... the number moves and then i click... its like turning up to a party when theyre stacking chairs...
so for me it doesnt mislead... it says the thing i already know in a way i cant argue with... you lot are talking about it lying to people who beat it... i just wanted to say theres a version where it tells the truth and the truth is im behind...
 
I have a former student to thank for a cautionary tale, one must consider that when a measure becomes a target it ceases to be a good measure, which is what is called Goodhart's law and which I used to illustrate with examination marks, and which applies with startling neatness to the closing line, he was a keen young man who had read that beating the close is the mark of a winning bettor, and he therefore set out to beat the close, which is a different undertaking from setting out to find value, he bet early in the week on every market he expected to shorten and took whichever side had been moving, and for a season his closing line value was admirable, nearly two per cent, he wrote to tell me so with some pride, he had in fact been paying the full margin on some four hundred bets in return for a price slightly better than a close that still had the margin in it, and his balance had fallen by about nine per cent, I wrote back to ask whether he was beating the close or beating the market, and I regret to say that he did not reply for about a year, I have myself recorded CLV in goals for the quarter-line markets and was embarrassed to discover that a quarter of a goal at the half-goal line is worth a great deal more than at two goals, so that movements which looked identical in my column were nothing like identical in value, a thermometer is no use to a man who has learned to heat the thermometer.
 
I recalculated this week against no-vig closes from the sharpest book in my records. The fifteen-year average moved from +1.9 points to +1.3. It is still positive, but it is not the figure I have been giving this forum.
Trust the process, not your gut.
 
It means it is unverified. Measure against the right close and wait for a few hundred bets.
 
One thing nobody has asked yet: what are your +1.4 "points" measured in? The same bet gives three quite different CLV numbers depending on how you report it, so it's worth knowing which one your sheet uses before you compare with anyone else's.

Example: you take 2.08 and the market closes 1.95 / 1.95.
- Raw CLV (your odds / closing odds - 1): +6.7%
- No-vig CLV (against the fair close of 2.00): +4.0%
- Probability points (fair 50.0% minus the 48.1% your odds imply): +1.9 points

Same bet, three numbers. At odds around 2.0, one probability point is roughly 2% of no-vig CLV, so if your +1.4 is probability points against a no-vig close it's about +2.8%, which would be a decent start. If it's against the raw close, part of it is just the margin.

On Klaus's point about longshots, the devig method matters too. The simple method oli described spreads the margin evenly and tends to leave longshots slightly overpriced; power or Shin put more of the margin on the longshot, so CLV on 4.00+ prices usually comes out lower with them. If you're unsure, use the most conservative result.

Disclosure: I'm the founder of Pinnacle Terminal. Our free CLV calculator shows all three readings side by side if you want to check your sheet: CLV Calculator - Raw and No-Vig CLV · Pinnacle Terminal
 

Similar threads

F
Early European Odds Value Betting - Beating the Closing Line Consistently
Replies
5
Views
1K
RedMerseyMind
R
Back
Top
GOALLLL!
Odds