This post is largely coming from the angle of statistical modeling however it would be of interest to see how sports fans react.

Pretty much as the title says, what do you do? How do you react?

When starting out, I was oblivious to data validation. However, over time data integrity has become a concern. At first, I noticed maybe 1-2% of match results would be incorrect for highly published matches and the results would only be out slightly, using soccer as an example: a 4-0 score would be erroneously reported as 5-0, the winner would universally be correct.

A minor inaccuracy, but nothing that would upset the fans, or damage the rapport with readers of a site and I would put it down as a typo / some other non-sinister reason.

For sports fan, a few might be irked by there team looking worse (the loser from the example above being made to look like they lost by an extra goal) but a loss is a loss and aside from the few fans of the loser in that case noticing... no harm, no foul.

For the mugs, if they happened to see the incorrect score, made a bet and lost... at the end of the day they were going to lose to the juice regardless. For modelers, 1-2% of total data having similar errors like that can eat into profits but if minor errors like that cause them to dip into the negatives, they should quit now while they are ahead.

Unfortunately, the rabbit hole is much deeper.

Now-on I tend to go over all new data with a fine-tooth comb. I still miss a lot but getting your hands dirty and not just blindly scraping into a fully automated database can be very eye-opening.

What I have found is that the 1-2% has remained unchanged for highly publicized global matches with say 10+ sources but tournaments limited to a region / sub-region eg Europe / East-Asia / Central-Africa is a very different story when there are only 2-3 sources. Many times there are tournaments with 5-20% of results being reported incorrectly which is insane. However the outcome (which team won) is usually still not affected and the inaccuracy usually amounts to a goal in soccer or a touchdown in American football.

Recently as I have been rebuilding the model and going OCD on data integrity... **** me. Any minor tournament or non-professional league with only a few sources should be watched like a hawk by all punters. Do not rely on your bookmaker to accurately grade games and there are too many unscrupulous / inexperienced ones out there.

The shark bookmakers may simply try shank there players and if coincidentally the result is reported inaccurately and you see that incorrect result then your chance to contest the result goes bye-bye. Alternately, some naive bookmakers in attempt to provide fast payouts may use those unreliable sources to grade their bets and chances are the punters are using those same site(s) as well because they want to know the result as fast as possible as well.

Unlike before when it might have only affected 1-2% of results, would not have affected Moneyline / 1X2 results and only some of the 1-2% spreads would have been affected, a good deal of punters could be positive for specific tournaments / leagues and some will be lucky to not be negative. As for the modelers, when the garbage-in gets that high, then the noise can run rampant in their models and it can easily cut profits in half or worse.

For the sports fans, how do you feel about this? Does this concern you or don't care? The latter is likely why the inaccurate scores remain in circulation.

What do other statistical modelers do / how do you manage data integrity in the specific situation below or similar situations?

- When adding data to model but it is not fresh eg 12+ months old
- No major / reputable score publisher reported results (eg FlashScore has no scores)
- Only 1-5 / very-few sources
--- Twitter and Facebook are inclusive of that count
--- Wikipedia has no results, only semi-finals / finals, or suspicious-results.
- No official scores available
--- Wayback Machine or similar has no archive.
--- Scores were taken off site, or...
--- Never bothered to publish... ****ing useless I know.
------ Instead they relied on news media outlets to "accurately" report results. No jokes, it genuinely happens way too often.


Most recent tournament that spurred this post:

- 17 scores + 1 missing score
- 10 scores: having conflicting sources (different scores)
- 5 scores: the result varies by the equivalent of 2+ goals in soccer
- 2 scores: the outcome changes for **** sake i.e. the winner changes

...

Looks like the post is going to fall on deaf ears... it is long, probably not new but certainly something very very rarely talked about on forums. Oh well...

Rather than spam up the forum with threads... if one of the mods / admins could PM SwoopAE...

"Are you from New Zealand? If not, I (user:tomsword) may able to point-in-the-correct-direction at the very least or otherwise help him/her with liquidity on a specific sport that he/she carved up with a double digit ROI recently."

Would rather not draw attention to the goldmine / particular-sport. Would also be interested in collaborating possibly but that is an off-forum conversation if SwoopAE is interested.

If you're going deep into historical data I can see how finding accurate results might be an issue, but i've never had a big problem in real time as bets are graded, yes flashscore etc make mistakes sometimes but if you verify through multiple sources i've been able to find/fix all of the misgrades that have happened to me (i'd estimate one misgrade every 500-1k bets in my case that the book doesn't catch/auto fix without me asking them to)

I'm not from NZ, i'm Australian for what its worth and sure, i'll hit you up to see if there's anything worth working together on. If you want to contact me off forum just tag me in something on twitter @capitao85 since you can't PM here and i'll follow you back and send a PM there.

@mods once toms has confirmed hes seen my post feel free to delete the thread as it's clogging up forum space

I was trying to make the post more neutral toward general readers as in from a punters perspective however based on the example at the end...
What would you do in terms of trying to get the accurate scores? Just keep looking and hoping, ask other modelers perhaps if they have the data, try talk to the news publishers (laugh), something else...

I am at the point where I am considering making a donation of AU$100-$150 to the official sports body for the official results. Though it just feels really wrong to have to pay for scores from an event that was even televised on SKY. Also I have never bought data and it is hard to trust an organisation will provide accurate data when they are the ones most likely propagating the false scores in a ******ed attempt to protect their valuable "copyrighted" facts.

Assuming that you might be getting double digit ROI on multiple sports and are unsure what sport I was referring. What is NZ really good at or used to be 3 years ago, other than shagging sheep? The TAB (in NZ) will typically take ~AU$8400 in cash bets at avg odds of $1.80 before they move the line. My bread and butter comes solely from that sport, so I prefer not to call attention to it. You may feel differently but it is truly a golden goose.

Also have you considered betting in cash at Tatts, they will take ~AU$1200 bets but will probably move $0.10, so you should be able to get the equivalent of AU$3600 on at avg odds of $1.80. Be careful though, if you **** Tatts anonymously, they will hire private investigators, rat you out to the ATO and may even talk to newspapers (though they may not publish a story). As well as spread your name around to whoever else eg other sportsbooks and the pigs. Unlikely they will do that for a minor sport / league but with enough persistence you could almost do enough to piss off the wrong person working there if you are unlucky.

More than a few K or $500 at Pinny as you stated in another post. Also even a marked account at Pinnacle that moves the odds $0.07 per bet should be able to get AU$1600* on average odds of $1.77 at them alone. Do not be put off by the OZ books, there is certainly not as much liquidity as NFL, NBA or EPL but there is still a lot and the market is insanely softer.

*Assuming 4x bets are placed just before closing at AU$400 a pop as well as that they actually publish lines and heaven forbid they do it a few hours in advance so the limits go up from ~AU$125.

Are you hitting Head-to-Head and Spreads only or Totals as well?

It's pretty clear you're referring to 7s (I don't beat any other sports for a double digit roi fwiw and theres a good chance my ROI isnt that high in 7s and I just had a good season although im very confident im winning at a decent clip over two seasons) - you don't need to keep dancing around it nobody on the forum is going to care at all

Yes, i've bet in person @ ubet terminals, although it's not very convenient and they often only take a few hundred at a time. You can do the same with tab terminals in other states although I haven't done so personally.

Mostly head to head and spreads, some totals but not as many

Can a mod give this guy PM abilities so that we can take this to PM this doesnt need to be a public thread as nobody else cares

Flashscore is pretty solid for results in real time as is bet365's live updates, for historical results you're right they can be difficult to find as it's a niche sport no one really cares about outside of Aus/NZ/SAF and a few smaller countries

I don't think you're going to develop a model to beat it based on previous seasons data unless you're trying to do totals and liquidity is worse for totals than sides. Obviously you def need accurate data if you're trying to build a model for totals.

I'm just finding obviously wrong openers or some teams that are over/underrated consistently based on form/squads/etc etc (eg Canada was consistently underrated last season in the lines) and if you watch every match in a tournament it can be pretty clear which teams are playing better/worse than usual or injuries happen and books don't account for them etc as books are using historical data to set their lines. Some books make another obvious mistake that is easy to exploit that i'm not going to talk about publicly as well.

Holy ****, this post is so long, who is ever going to r--






Oh.

@HolidayInTheSun: That is why there is a question in the second paragraph, perhaps it should have been prefixed by TLDR.

@SwoopAE: I have only been watching the Australian situation / market from a distance though over the last year but doubt much has changed.

TABcorp is a waste of time from a 7s perspective, they are far too slow to put up odds to be able to put cash bets on if juggling (comparing) books unless trying to completely saturate market.

Tatts appear to be posting consistently at least ~6 hours in advance for matches 1-8 (day 1) and 25-32 (day 2), and many times posting Totals which is significant due to their higher relative limits to elsewhere.

Already built model and used, rebuilt and used many times over, the ROI of even the earliest serious iteration is simply not believable so it is not worth publishing for both Spreads-vs-any-book-but-TAB-NZ and Totals-vs-any-book (will only solicit trolling but you know yourself how soft the markets are).

Your double digit ROI is not from favourable noise in results. Half is likely due to Bet365 shading pulling the market out if you are using some other book's line for performance or if using Bet365 lines that they are shade-aholics, and the other half is you outsmarting them. In fact, the shading is much less than it was 3 years ago so it is possible that only a quarter is due to shading these days. I take that back if you are achieving 20+%, but a not insignificant portion is still due to them shading.

The wrong scores published has been an issue for internationals but what spurred post is currently researching regional club level. Data gets very scarce and unreliable at that level, worse than trying to find amateur junior tennis results for new kids in the ITF from non-English countries.

Opening TAB-NZ spreads is a no-go (or was). However throw some Bet365 shading into the market and TAB-NZ slowly moves out, they can be beaten but it is not for the faint hearted. So that AU$8400@1.80 at the TAB-NZ does come with caveat that you are playing against some of the sharpest (initially) in the rugby markets. However they are very arrogant which is why it is possible to get down 2 large bets that any other book would be scared ****less taking one of before they move the line back in. The trick is that they will take that sort of money on Totals as well yet...

TAB-NZ are clueless when it comes to Totals in rugby. They copy Bet365, so easy to beat lines and high limits hence no liquidity issue. I do not know why they do not hire a modeler or two to brush up that area and they could really dominate. In fact I do not know how anyone can even have a model that outputs a handicap result only yet there are hundreds of posts from sharps over the net which seem to corroborate / imply handicap only models. Spreads and Totals are co-dependent. I am sure there excuse for Totals is weather but seriously how hard is it to look out the window, check a live weather cam (if outside NZ) or pay for a professional forecast service.

Have some comments on injuries but will save that for later.

Tom, let this thread die, just message me on twitter @capitao85 as we can PM there or i'll message you my Skype username or whatever there if you want to talk 7s, there's no reason it needs to be a public conversation on a forum when no one else cares about it. I am interested in discussing it with you as we're two of the only people I know interested in such a niche spot but this isn't the place for it. Obviously we both know 7s is absurdly beatable, but most of the guys here are US/Euro based and have no interest in it due to the low limits. I do have more outs to get action down this coming season than I did last season and i'm happy to discuss with you privately.

@HITS - The battle had been lost, or so they claimed as something something on the normandy beaches. Something something Paul Casey, and they lived to fight another day.

Honestly if I hadn't had a recent experience I would have completely written this off as quackery, but something happened to me recently.

A book I've used for YEARS and has never had a problem made a HUGE misgrade in my favor. I had a 4 leg parlay to win $4k and they graded it as a win. However, one of the games lost, it was a 2nd half of a ncaa football game, a division 1 game between 2 major teams, and i had over something...I wouldn't have even known because I was on such tilt that I went to sleep, and woke up in the morning, saw my balance and realized I won. However, I got a funny feeling and for some reason just checked the previous day's scores. The 2nd half had stayed under by less than a touchdown. So possibly something like this affected it, but very strange that it was in my favor. I cashed immediately out of that account and haven't used it since.

Ive had a few issues too, especially player performance bets.

One bookie I use terms state, that they will grade at the box score that is current at end of game. Probably 1 in 300 are wrong.

Flash score can be quite bad for rugby, maybe 1 in 100 times ive seen the wrong name come up for the try scorer.