Any opinions on sites like Team Rankings, Prediction Machine, Accuscore, & Numberfire

[Vampiro]

Any of them worth subscribing to as a supplement and or extra filter when handicapping and narrowing your card down?

Would like to maybe pick one of them to use for Baseball and maybe even NHL & NBA. Since those three sports have such long seasons and more of a daily grind than football season.

I apologize if I put my question in the wrong forum. Thanks in advance for any opinions or suggestions.

[nash13]

I tried to create angles by combining the predictions of masseyratings, dunkel index and dci. worked great for me, but it is very very very hard to do so on a daily basis with manual skills.

[Vampiro]

So nobody here is really all that familiar with any of the four I mentioned.

[tb1984]

Any of them worth subscribing to as a supplement and or extra filter when handicapping and narrowing your card down?

Would like to maybe pick one of them to use for Baseball and maybe even NHL & NBA. Since those three sports have such long seasons and more of a daily grind than football season.

I apologize if I put my question in the wrong forum. Thanks in advance for any opinions or suggestions.

I think you should post under the "Service Plays" forum http://www.sportsbookreview.com/forum/service-plays/

[Vampiro]

I think you should post under the "Service Plays" forum http://www.sportsbookreview.com/forum/service-plays/

Thanks for the advice. Wasn't really sure if I asked in the right forum or not.


MODERATORS if one you could move this to the right forum for me. Thanks in advance.

[evo34]

Definitely not the wrong forum -- those sites all provide data as well as picks, and the discussion is about their utility, not promoting/distributing their plays. I have used Team Rankings and Accuscore. Bluntly, Accuscore is dishonest (they calc. all NHL totals picks with -110 juice, for example), and I have no evidence they have any skill. Team Rankings used to have a nice edge as recently as three years ago, but that's long gone. There are some useful data on that site for free, though, so I do recommend checking that part out.

[LT Profits]

Definitely not the wrong forum -- those sites all provide data as well as picks, and the discussion is about their utility, not promoting/distributing their plays. I have used Team Rankings and Accuscore. Bluntly, Accuscore is dishonest (they calc. all NHL totals picks with -110 juice, for example), and I have no evidence they have any skill. Team Rankings used to have a nice edge as recently as three years ago, but that's long gone. There are some useful data on that site for free, though, so I do recommend checking that part out.

Do you prefer TeamRankings or kenpom for CBB? Agree that TR is great for other sports data wise.

[snark]

Kenpom is by far the best for CBB.

[evo34]

Do you prefer TeamRankings or kenpom for CBB? Agree that TR is great for other sports data wise.

I don't do much fundamental analysis in CBB, so no view there. Funnily enough, TR has a couple baseball stats that are tough to find elsewhere.

[evo34]

Kenpom is by far the best for CBB.

I assume you mean for building your own ratings, not for using his power ratings, right?

[Slick Lizard]

TeamRankings is the only one of the four that were mentioned that I use regularly. As mentioned, AccuScore is not worth the time to look at much less paying any $$$. Prediction Machine and Number Fire are ok at best IMO. As for TR, the site is a tool to use, I wouldn't just take their numbers and roll with it but the site is full of info that is useful for someone that knows how to use it.

[trader4046]

Numberfire

TeamRankings is the only one of the four that were mentioned that I use regularly. As mentioned, AccuScore is not worth the time to look at much less paying any $$$. Prediction Machine and Number Fire are ok at best IMO. As for TR, the site is a tool to use, I wouldn't just take their numbers and roll with it but the site is full of info that is useful for someone that knows how to use it.

I have used numberfire for the last 2 months mostly for fantasy help but I just really started betting with the NBA over/unders from their site. In the last 2 weeks of betting only o/u at 3* or above they are 34-9.

[TheMoneyShot]

I use TeamRankings time to time for Home/Away differentials. Pretty cool site.

[Bsims]

Interesting topic. My opinion is that none are probably worth he money. However, by coincidence, I decided to do a study involving a couple of these (KenPom and Numberfire) to see if one was consistently better than another. Then I'll throw in a couple of my approaches. It does look like TeamRankings may have back simulated scores available without subscribing, so I might add them. I'll probably post an update in a few weeks.

[Bsims]

I have concluded the study I mentioned in the previous post. I focused primarily on CBB because I had access to data from both KenPom and NumberFire. I did this over largely a two month period, December and January. I built a file with final scores and predicted scores from these two sites. I added the predicted scores implied by the LV lines and threw in two of my old systems. I then did correlations on the actual to predicted scores. My assumption was that to be valuable, the predicted scores from the sites would have to have a higher correlation to the actual scores than did the implied LV scores.

As expected, they did not. Over 3,041 CBB games, the implied LV scores had a correlation of 0.594. KenPom's correlation was 0.571 and NumberFire's was 0.542. The results were similar for NumberFire in the other 3 sports, lower correlation than the implied LV lines. Hence, I don't believe these are worthwhile.

The original post also asked if these could be used as a filtering device. While there are several approaches to doing this (I looked at regression analysis using several of the projected scores), I concluded they would not be worth the effort.

As a side note, rather unfortunately, one of my old systems I threw in had higher correlations than any of the others and would seem to provide a worthwhile wagering vehicle. I say unfortunate because I apparently last did analysis on this approach 14 years ago, and now feel compelled to re-analyze this approach. This will require a lot of work and likely not lead to anything productive.

[RockBottom]

CBS Sportsline is offering a similar service now.

[HeeeHAWWWW]

My assumption was that to be valuable, the predicted scores from the sites would have to have a higher correlation to the actual scores than did the implied LV scores.

As expected, they did not.

Yeah, that's pretty much what I'd expect too of any public rating system.

[d_mad]

Surefire and Accuscore come with a ESPN subscription. They never seem to accurate. Doubt there is much value to them.

Also a couple times I have seen where they don't adjust their simulation for injuries.

[bookie4ever]

Anyone having problems getting to Teamrankings? Im being redirected

[zlaroc]

Warriors !!!

[statnerds]

read through this thread 3 times and, by golly, i didn't find a single thing in here that the isn't already in the price.

let the market do the work for you.

[Alpha male]

bump

[Roscoe_Word]

Bsims analytical guru.....hope all is well.

[Alpha male]

Bsims analytical guru.....hope all is well.

dafuq is that?

[Alpha male]

bump

[Roger T. Bannon]

There are no free rankings or subscriber ratings that will work for betting against the line. They are a little better than the AP ratings but that is about all. The only thing they are really good for is projecting what a line might be.

None of those listed in this post were particularly good.

The best ratings are probably ESPN FPI, ESPN SP+, Massey Peabody and Dokter

To beat the line, you need something more baked in that these types of products offer but they are pretty good so far as rankings go.

When it comes to other sports other than college football, they would be even less valuable.

[Roger T. Bannon]

The basic problem with this theory is that you if you use a ratings as your "starting point," how do you get to the end point?

[Roger T. Bannon]

I'll share a little experiment I did with this sort of thing. I had a theory that if you took the very best bet from the very best off the shelf models and bet them each week, it might be profitable. So I determined the five best college football rankings over the previous 5 and 10 years and took their best bet. The line on these games was off at least 7 points according to the rating.

The first year the picks hit 60%. Over the next few years the win rate dropped to 52.5% overall. At that point I quit tracking.

I did the same sort of thing with college basketball and quickly found that there were few college basketball ratings and the lines were rarely off by more than 2 points. This is because they use the line in their ratings. The picks hit about 45%. By doing this I got familiar enough with CBB that I started making my own bets and hit about 56% for the rest of the season.

[Roger T. Bannon]

The first year the picks hit 60%. Over the next few years the win rate dropped to 52.5% overall. At that point I quit tracking.

Looking back, I actually quit at the end of year 2. All of the outperformance in Year 1 came from just 1 rating that only lost 1 or 2 games. That was just a fluke.

[Bsims]

As a side note, rather unfortunately, one of my old systems I threw in had higher correlations than any of the others and would seem to provide a worthwhile wagering vehicle. I say unfortunate because I apparently last did analysis on this approach 14 years ago, and now feel compelled to re-analyze this approach. This will require a lot of work and likely not lead to anything productive.

The resurrection of this thread led me to search my historical files and documentation to identify which of my old systems showed some potential. Once again, the search proved fruitless. I couldn't find any find any information from about 20 years ago that would indicate even what system that was. Too many years, systems, and medical interruptions. Sure wish I had done a better job of documenting. Let that be a lesson to us all.