Capping and Discussions, 4/2

**Illusion** · 04-01-06, 09:51 PM

Bump for Razz.

**Razz** · 04-01-06, 10:09 PM

Haha. I'm ready to see the Edgar line, because there is an NBA game I am thinking about playing tomorrow.

**mlwitchking** · 04-01-06, 10:10 PM

Originally posted by Razz

Haha. I'm ready to see the Edgar line, because there is an NBA game I am thinking about playing tomorrow.

where does the "edgar line" come from/rely on?

**Jay Edgar** · 04-02-06, 12:24 AM

MWK , it's based on power ratings from the last 10 games only. The best bets thread from Feb 27 (or 28?) has a longer description.

Gotta wait for these dog-ass Blazers to finish their latest loss before it's finalized. Have it in a few minutes.

updated:
===================
NBA Sunday

PHI (#28) by 6.5 over NY (#29)
DAL (#11) by 4.0 over DEN (#12)
DET (#5) by 4.6 over PHX (#7)
MIN (#18) by 5.8 over GS (#22)
LAX (#6) by 6.4 over HOU (#23)
MEM (#14) by 7.0 over ATL (#26)
CLE (#10) by 1.0 over CHA (#24)
NJ (#1) by 7.2 over MIA (#4)
TOR (#21) by 4.2 over NOK (#27)
SAC (#15) by 3.7 over LAC (#16)
SEA (#20) by 10.2 over POR (#30)

**mlwitchking** · 04-02-06, 12:42 AM

i didn't find a longer description in the feb 27 post. looking forward to your line though.

**Jay Edgar** · 04-02-06, 01:40 AM

Originally posted by mlwitchking

i didn't find a longer description in the feb 27 post. looking forward to your line though.

(warning -- RIDICULOUSLY long post ahead -- won't hurt my feelings a bit if any brother decides to skip it)

MWK, maybe it was Feb 28 -- I'm not 100% sure without checking. Sorry about the wild good chase.

And actually this might be a good time to explain/improve the concept as I am thinking about trying something similar with BASEBALL -- and am very interested in any thoughts or similar experiences.

(Although with baseball I know I'd need a team offense power rating, a starting pitcher power rating, and likely a bullpen strength/availabilty power rating, Then we'd need to mix them and come up with a money line and, I hope, a total. I actually think that this tool would have a better chance with baseball totals than NBA totals because in bases skill=scoring much more directly, unlike with baskets where you need to worry about pace and style of play)

The pointspreads that I have been putting up nightly are based on the estimated gaps between teams of different strength levels. To estimate those gaps, I looked at five years of winning percentages and margins of games and did a formula to derive a pointspread from the difference bewteen any two power ratings on a 0-100 scale.

To get the power ratings to plug into that pointspread formula, I look at each team's last 10 games. In each of those last 10 games, each team has been graded with a power score. The 10-game power rating is the weighted average of the team's power scores over the last 10 games.

In giving each team a power score for every game, I use a formula that looks at the final score of the game, where it was, and how strong the teams were coming in.

I'll give an example, and I'm going to make up some numbers because the formula numbers are embedded in the spreadsheet -- I knew them when I wrote the formula, but now I've forgotten them.

Today, for example, CLE beat MIA by 7 at home. Adjust the score (106-99) to the league average of 186.3 points, and CLE scored something like 51.7 of the points and MIA 48.3 of the points. Those are their base game scores. But CLE's effort came against the #3 team in the league (and over the last 5 years the 3rd best team in the NBA wins at a .605 clip on the road) so there's a big boost above 51.7 there. MIA lost, but the loss came to the #12 team in the league (which wins at home at .668 clip over the past 5 years) , so there's a boost above 48.3 there. When the complete formula is applied, CLE got a score of 63.3 for a very good win, and MIA got a score of 52.7 for a loss that was not such a bad loss.

So to get a team's power rating, you calculate their last ten game scores (and right now the most recent game score is multipled by 10, then the one before that 9, and so on -- and then the whole thing divided by 10+9+8+7+6+5+4+3+2+1 -- although I am thinking of tweaking that formula because it puts just a little too much weight on the very most recent games).

I may be confused by my own stuff, but I think of the power ratings (that's the list coming up just below) as a team's projected winning percentage on a neutral court against a dead-average opponent. ( I have not yet tried to use the ratings to get a reliable money line, however.)

Just to flesh things out since this is already an absurdly long post -- going into tomorrow, the results of each team's last 10 game scores yield this:

Power ratings as of Apr 2nd, AM:

1NJ 72.6
2SAN 70.7
3WAS 60.7
4MIA 59.6
5DET 58.7
6LAX 57.5
7PHX 56.9
8UTH 56.7
9ORL 56.6
10CLE 55.4
11DAL 55.2
12DEN 53.9
13MIL 53.6
14MEM 53.4
15SAC 52.8
16LAC 52.3
17BOS 51.7
18MIN 50.9
19CHI 50.3
20SEA 46.6
21TOR 46.1
22GS 44.5
23HOU 44.2
24CHA 42.5
25IND 42.3
26ATL 41.3
27NOK 39.9
28PHI 37.2
29NYY 23.7
30POR 22.3

(The average of these numbers is 50.3, and it has been about 50.3 all year. I imagine that, for what I am trying to accomplish, in a perfect world the average would be 50.0, but I find 50.3 encouraging enough)

(Also by the way, the distances by which 1-2 and 29-30 are away from the pack are huge -- bigger than they have been all year. On a neutral floor tomorrow, this tool would favor #1 NJ over #30 POR by 15.5 points. On February 1, just to pick a date, SA was #1 at 68.4 and CHA was #30 at 32.2, with the Spurs favored on the neutral court by 11.1.)

And so for example with the ATL-MEM game tomorrow, MEM's power rating of 53.4 and ATL's of 41.3, plugged into the formula, would make MEM a 3.7 point favorite on a neutral court. I give them 3.3 for the home court (using the average of two homecourt estimates on the internet, not my own calculation)

That makes MEM, ranked 14th in the league, a 7.0 favorite over ATL, ranked 26th in the league.

How would this work for baseball? Well, I'm thinking about that. I know that if, I had unlimited time, the tool at the core of it-- what I would want to rate the league's teams and pitchers on their ability to produce (teams) or limit (pitchers) -- would be expected run value. But that's a subject for another post.

**Dark Horse** · 04-02-06, 06:48 AM

Jay Edgar, do you use different league averages for different months?
I used 181 earlier, but I know it goes up during the season, and your 186 shows by how much. For totals it may be good to have a monthly, or even biweekly, graph for the league average. For lineoffs it doesn't matter.

**Wassymac** · 04-02-06, 07:03 AM

I got on the same sides as RAZZ while I was still in England as soon as I saw the WNCAA lines come out. UNC wins by double digits. LSU wins straight up.

**Jay Edgar** · 04-02-06, 09:16 AM

Originally posted by Dark Horse

Jay Edgar, do you use different league averages for different months?
I used 181 earlier, but I know it goes up during the season, and your 186 shows by how much. For totals it may be good to have a monthly, or even biweekly, graph for the league average. For lineoffs it doesn't matter.

DH, I haven't tried to pick totals based on number-crunching, but I'm sure you're right that you'd want to look at whether the median or league average was on the move, or was different in different times of the year. (Especially if there were recent rule changes or some other likely explalantions for a departure from a past pattern.)

Also as you suggest, a hyper-accurate average or median league total points isn't necessary to what I'm doing. I'm just looking for some common scale on which to reset the 74-64 win and the 141-131 win so that the ratings don't count them as equally decisive, because I don't think they are.

**mlwitchking** · 04-02-06, 04:43 PM

i don't understand how cleveland is only a 1 point favorite against charlotte in your rankings. one team has been tearing it up lately, while the other is doing alright.

**mlwitchking** · 04-02-06, 04:48 PM

As for your weighting of the recent games, how about this:
15%,14%,13%,12%,11%,9%,8%,7%,6%,5%

The last five account for 65% now, instead of 72.7% and the latter games account for more.

**mlwitchking** · 04-02-06, 05:04 PM

i feel like an idiot for not picking golden state at +8. there is no reason for minnesota to be -5 or higher against anyone.

**The Great One** · 04-02-06, 05:11 PM

Not a flame on anyone, but has anyone actually watched these teams play before putting $$ on them.

I hate girls basketball. I think it should just be outlawed. I'm of the thinking. If they are that good, play with the men. i just hate how they try to act like they are athletic and the truth of it is, most 8th grade boys players could school professionals.

And I actually have proof to back this up. Its hard for me to come up with proof on my "I'm a better athlete than babe Ruth" thing.

But, I've played a WNBA player one-on-one a few years ago and destroyed her. Granted, she was in college at the time, but it wasn't even close. I think she was a 4 year starter at a major university that did win a national championship when she was there and still plays in the Girls league if I remember.

The reason girls shouldn't count is the way they shoot. They have to push the ball away from their

everytime they shoot. If coaching girls basketball, just concentrate on rebounding and lay-ups cause thats all you need.

**Jay Edgar** · 04-02-06, 06:03 PM

Originally posted by mlwitchking

As for your weighting of the recent games, how about this:
15%,14%,13%,12%,11%,9%,8%,7%,6%,5%

The last five account for 65% now, instead of 72.7% and the latter games account for more.

That's a good thought.

The other thing I might like to do is weight the most recent game no more heavily than the game or two leading up to it. The market overreacts to the most recent game very often -- I'd like to avoid the same mistake. So maybe something like; 13-14.5-14-12.5, then through the final six on the values you suggest.

**moses millsap** · 04-02-06, 09:15 PM

I thought for sure LSU's 45 against UCLA would have been lowest total of the weekend, but the LSU women look even worse, lol. 4 points in 9 minutes plus

**Alamorich** · 04-02-06, 09:19 PM

Looks like UNC and LSU players started their periods on the same day.

**Razz** · 04-02-06, 10:15 PM

Originally posted by OWNED

I thought for sure LSU's 45 against UCLA would have been lowest total of the weekend, but the LSU women look even worse, lol. 4 points in 9 minutes plus

No shit. At least they scored more points in the first 8 minutes of the second half than they did in the entire first half.

**mlwitchking** · 04-03-06, 12:30 AM

personally, i think the last five should count for about twice as much as the five to ten previous games. i also agree that the market reacts to the latest game as well. on a side note, i am thinking of fading the nets soon, because, alas, all good things must end.

**Dark Horse** · 04-03-06, 06:29 AM

Originally posted by The Great One

Not a flame on anyone, but has anyone actually watched these teams play before putting $$ on them.

I hate girls basketball. I think it should just be outlawed. I'm of the thinking. If they are that good, play with the men. i just hate how they try to act like they are athletic and the truth of it is, most 8th grade boys players could school professionals.

And I actually have proof to back this up. Its hard for me to come up with proof on my "I'm a better athlete than babe Ruth" thing.

But, I've played a WNBA player one-on-one a few years ago and destroyed her. Granted, she was in college at the time, but it wasn't even close. I think she was a 4 year starter at a major university that did win a national championship when she was there and still plays in the Girls league if I remember.

The reason girls shouldn't count is the way they shoot. They have to push the ball away from their

everytime they shoot. If coaching girls basketball, just concentrate on rebounding and lay-ups cause thats all you need.

Congrats. The MCP award, without a close second in sight.

A double victory, because you've also made my ignore list.