Does More Research Ever Make You Worse?

[CoachTony_Bets]

Last Saturday I watched nine hours of film on one game. Both teams, three weeks each, tagged every third down. I felt great about it. I placed the bet and it lost by a point on a fumble, and that isn't the part that bothers me. The part that bothers me is that I looked at my log afterward and the games I spent two hours on and the games I spent nine hours on have about the same record.
As a coach I know this from the other side. We cut the playbook every Tuesday, because Friday's version has to fit in a sixteen-year-old's head under pressure. A kid with forty reads on one play has none. The good ones have three. I think there's a middle ground, but I've never found where it sits for betting.
So I'm asking. Is there a point where more research makes you worse? And do you know where yours is?

 

[oli_sussex]

Yes. My model has seven inputs. At twenty-two it did worse out of sample, so I removed fifteen.

 

[patchworkfox]

Last Saturday I watched nine hours of film on one game. Both teams, three weeks each, tagged every third down. I felt great about it. I placed the bet and it lost by a point on a fumble, and that isn't the part that bothers me. The part that bothers me is that I looked at my log afterward and the games I spent two hours on and the games I spent nine hours on have about the same record.
As a coach I know this from the other side. We cut the playbook every Tuesday, because Friday's version has to fit in a sixteen-year-old's head under pressure. A kid with forty reads on one play has none. The good ones have three. I think there's a middle ground, but I've never found where it sits for betting.
So I'm asking. Is there a point where more research makes you worse? And do you know where yours is?

It can, mostly through confidence rather than accuracy. Past a point extra hours stop changing your read of the game, but they make you feel surer, so stakes go up and you are less willing to pass when the price is wrong.

Win rate is also the wrong yardstick here because it needs a huge sample before two groups of bets look different. A faster test: log the price you took and the closing price for the two-hour games and the nine-hour games. If the long-research bets do not beat the close more often or by more, the extra hours are not finding anything the market has not already priced.

And a lot of what moves a line late is information you cannot get from film anyway: injury news, lineups, weather. Most edges come from a few inputs that matter, not from forty reads.

 

[SharpEddie47]

The question needs one distinction, or this thread will argue past itself. Research that finds information the market does not yet have is valuable. Research that rediscovers information the market already has feels identical from the inside and costs the same hours. Past a certain point, nearly all research is the second kind.
I can put a number on where that point sits for me. For 300 bets I logged the minutes spent on each, along with my stated confidence. Bets researched for under 30 minutes: 210 of them, +4.9% ROI. Bets researched for more than 90 minutes: 38, +2.8%. The gap is well inside noise, so I do not claim the longer research was worse. I claim something more modest: I cannot detect that it was better.
The confidence column is the interesting one. Average stated confidence rose from 58% in the short group to 69% in the long group. Hit rate did not move. Ninety extra minutes bought me eleven points of certainty and no accuracy.
That is the mechanism. For most people, more sure feeds the stake. It does not feed mine, which never moves, but it does feed my willingness to place the bet at all.
Trust the process, not your gut.

 

[FadeThePublic]

Here's what nobody wants to hear: my research is mostly lawyering. I decide, then I research. The seventh tab I open is always the one that agrees with me. I used to run six betting percentage sites at once. When they disagreed, which was constantly, I trusted whichever one matched my pick. (Peak sharp bettor: a man with six instruments who only reads the one that's already nodding.)
The fix was almost embarrassing. One source for public percentages. One for line movement. I'm not allowed to open a second. If the one source says I'm wrong, I'm wrong. RIP public money, but also RIP my ability to find a reason for anything.
The research didn't make me worse because it was bad. It made me worse because it was a buffet.

 

[TaffyTipster]

Mine was the pub.
Team news in the pub is five blokes who all know somebody.
"Heard he's out."
"My cousin says he's fit."
"Saw him at the garage."
By the time I'd heard five I felt like I'd researched it.
I hadn't.
I'd collected five opinions off men who'd each heard it off one man.
Never wrote any down.
Never checked any.
Bet on it anyway innit.
Calling it research made it feel like work butt.

 

[Klaus]

The question is imprecise. More research does not make a bettor worse. More of a certain kind does.
There are three kinds, in descending order of value. Information the market does not have. Information the market has but has priced incorrectly. Information the market has and has priced correctly. The third is most of what is available, and it has zero expected value.
The difficulty is that all three feel the same to the person reading them. I therefore ask a narrower question of any new input: does it change the model output by more than the model's own error? If it does not, I do not read it.

 

[ThePuntingProf]

I keep a column in the ledger headed hours, which began as a device to discourage myself from doing too much and has become a record of precisely the reverse, one must consider what it shows when sorted, the Arsenal fixtures, which I have mentioned before as my least profitable category, received on average six hours of preparation each, the other matches received an hour and forty minutes, I had always supposed that the Arsenal figures were poor in spite of the care I took over them, and it did not occur to me until this month that they might be poor because of it, the extra four hours were not spent discovering anything, I know this because I can read my own notes from those evenings, they consist of assembling a case, every injury is weighed in the home side's favour and every poor performance receives an explanation, when one loves the subject every fact becomes a friend, whereas on a Tuesday night in the second tier about which I knew only the table I had no case to build and simply followed the numbers, and those matches returned the best figures in the book, I am obliged to conclude that the hours I valued most were the hours that did the most harm, and that knowing a great deal about a team and being in a position to price it are quite different attainments.

 

[ParlayPrincess_88]

Okay so I did an experiment lol. Because of the holdout thing I flipped a coin for every game. Heads = research a LOT (like 45 minutes), tails = barely (5 minutes). Six weeks.
Heads: 22 bets, 9 won. Tails: 31 bets, 15 won. So 41% vs 48%. Which means research makes me WORSE?? or I'm just bad lol. I don't know how to read it.

 

[SharpEddie47]

Princess, the design is good. Randomization is exactly the right instinct. The result is not usable. At those sample sizes, 41% and 48% are well within what two identical processes would produce by chance. The honest conclusion is that you cannot tell. You would need several hundred bets per arm.
Trust the process, not your gut.

 

[DublinDegen]

lads the research was never about the bet for me... id spend two hours reading form and injury news and stats and feel like i was doing something serious... and then id place a bet on a completely different game... in play... because the game id researched was already priced and boring...
the research was the part that let me feel like a person who does this properly... a ritual... the way some people light a candle...
so for me more research made me worse because it was never research... it was permission...

 

[ParlayPrincess_88]

Nine HOURS of film?? I can't watch nine hours of anything except Chiefs lol

 

[CoachTony_Bets]

It's less fun than it sounds, Princess. Mostly guys standing around between plays.

 

[CoachTony_Bets]

Eddie's confidence number is the film problem. Nine hours made me feel like I'd earned the bet. The bet didn't get any better. I just got louder in my own head.

 

[SharpEddie47]

Fade, the issue is independence, not quantity. Six betting percentage sites that scrape the same data feed are one source, not six. Do the math: your disagreements were probably rounding and timing.

 

[FadeThePublic]

...So the buffet was one dish. (Great. Six plates, same dish.)

 

[Klaus]

Eddie is correct. Six sources carrying one underlying number have the information of one and the appearance of corroboration of six.

 

[oli_sussex]

Confidence rises with hours. Accuracy does not. Same in my logs.

 

[TaffyTipster]

Ten seconds to say no to something, Klaus mate?
I'd need a pint first.

 

[Klaus]

Ten seconds to reject is cheaper than forty minutes to accept.