Attempt to quantify SLU 2nd half results using BPI

[GW]

I was interested in quantifying SLU's 2nd half results. The 2nd half team is a seemingly far different entity from the first half team, through no fault of their own (Majerus, injuries). My belief is that SLU put the first half season behind them and perhaps even used it as a motivation device to become the elite team they were meant to be all along. I had requested of the wiz to see if he could isolate second half SLU numbers to give us a clearer picture of this developement. Unfortunately, it would require too much tweaking of the program he uses to provide these stats. I decided to see what I could produce on my own.

Disclaimer- understand that I am not anywhere in the league of the wiz, nor claim to be a stats guru of any sort or intend to become one. Just doing what I can to fill in the gap for informational purposes only....

I decided to use epsn's bpi rankings. These are similar to the wiz's in the sense that they incorporate margin of victory into their calculations- it doesn't just matter that the game was won or lost, but by how much you won or lost is critically important. Are you winning by getting by with the skin of your teeth in overtime or are you destroying your opponent by 30 points in regulation? Bpi incorporates other things into their calculations as well (I don't know if they really advertise everything they put into it...) but it serves my purpose in the sense that a ranking is given for every specific game and thus whatever time frame someone desires can be calculated.

SLU's Bpi number at the end of the season was 80.9 ranked #19. Calculating the Bpi based on SLU results post duquesne produced a value of 93.3 which would be ranked #1 on a full season basis (Louisville is #1 with a 91.3 value). The early season Bpi based on SLU results from duquesne and earlier produced a value of 71.6 and ranking of #56 which is about where Temple is. So according to Bpi, in the first half of the season we were playing Temple quality basketball and in the second half we have been playing to a level exceeding that of Louisville (full season). I was suprised by these numbers- its readily apparent that the team has been reborn if you will this second half but that is a vast increase. Just imagine if the whole season could have been played at this level... Another interesting thing is that when comparing the pre and post duq teams against each other, bpi projects post duq SLU beats pre duq SLU by about 14 points- which actually seems about right considering....

Another interesting calculation I made was SLU's Bpi post duq in "games that mattered most". For these I included vs top 25 teams, playing for conference reg season title (lasalle), playing for conference tournament (charlotte, butler, vcu), and against New mexico state. SLU is 8-0 in those games with a Bpi value of about 96.8. This is very good to see as it says that SLU is rising to the occasion in the biggest games- bodes well going forward.... That 96.8 value against Louisville's full season value of 91.3 would give you a projection of SLU by about 5 points. While this is a good result to see- if we are going to make adjustments for SLU we need to be fair to Louisville as well. Louisville has been playing lights out starting around the same time frame that our surge began. Louisville had a 3 game losing streak mid season but since then, they like us have a single overtime loss on the road. From the Pitt game forward, Louisville's Bpi value is 95.7. That then gives a result of SLU by 1 if we play at ultra high "games that matter" level, Louisville by about 2 if we play at average second half standards. To find that SLU may be realistically considered to be about even with Louisville is very encouraging... When just eyeballing all the other top seeds, no other team has Bpi values anywhere close to where SLU and Louisville's are (2nd half). So if SLU does get there that game becomes key as whoever wins will be favored by at least a few points in all games to follow... (granted we are viewing SLU in the most favorable light possible, disregarding the first half results, but given the circumstances I dont think it's unrealistic to do so....).

Concerning the SLU vs. Oregon matchup, Oregon has a full season bpi value of 76.9 ranked #33. Bpi has a method for adjusting when teams are missing a top player so I'm disregarding the 5-4 results when they were missing Artis. This adjusts their Bpi up to 79.5 which is ranked #23. The post duq SLU bpi value of 93.3 vs Oregon at 79.5 gives you a projection of SLU by 10 points. Using the ultra high SLU ranking for "games that matter" gives a projection of SLU by 13 points. It is important to note though that the last 3 games Oregon has played about its best of the season. For the last 3 games Oregon has a Bpi of about 97.1 If that high level Oregon team is the one that shows up its about even with our "games that matter" team. I don't know if 3 games is enough to identify a trend but its worth mentioning. I think its important for us to make Oregon aware as early as we can in the game that they are actually the #23 overall team and not the ultra high level team they are now playing at- expose the last 3 games as an anomaly rather than a trend. I'm hopeful that all of their righteous indignation at their admittedly poor seeding was discharged at the expense of oklahoma state. If they got too emotional about it they may have "spent their wad" so to speak...

I estimated SLU's Bpi value against New mexico state to be a 98. I know the game looked easy at times but that is a testament to how good SLU is playing right now. New mexico State is decent- has a bpi value of 66.4 which ranks #74. They are just below Xavier (#72) and just above Umass (#79). So it was like beating Xavier or Umass by 20 at a neutral site... One final thing- I know that Louisvile taking down NC A&T may have appeared more impressive than our win as it was by a much larger margin (31 points vs 20). NC A&T has a bpi value of 39.8 ranked 236. They had to win the play in game 2 nights earlier and it was a de facto home game for Louisville. Applying only 2 points for home court advantage (rather than a standard 3 or 4), and not considering the effect of the extra game, Louisville's bpi for that game is about 98 as well. So basically, they were about equal performances by the two teams who are currently playing at the highest level.

NOTES:

I performed one individual tweak on the Bpi calculations in regard to how they account for overtime games. Looking at the numbers, they base their calculations on what the final score was as opposed to viewing it as a tie since the two teams played to even through regulation. This is important because if you lose by 50 points in overtime Bpi calculates the result as if you had lost by 50 points during regulation- no difference. My belief is that this unnecessarily "penalizes" you for getting to overtime- you could have lost by 49 during regulation and Bpi would give you a slightly better grade for this. Had you lost by just a point during regulation you would have a night to day vastly better grade. So my view is that overtime results should not be considered in the basic calculation- it should be looked at as a "tie". This is important for SLU because the 11 point overtime loss against Xavier is viewed as an 11 point loss as opposed to a tie. The Bpi value for that game considering it to be is 11 point loss is 43.6 which is the the grade just below CS northridge. Considering it a tie gives a value of 75.2 which is the season grade between butler and stanford. In other words, my adjustment views it that we played a butler quality game that night (was just a slightly off night, which fits with what happened...) instead of saying we played a CS northridge quality game (ouch...). Now if you wish to argue that the overtime results should still be counted then I say worst case the basic numbers should include an asterisk- that they are only based on results through regulation. Now if the game were to go to overtime, ya maybe if 4 of our guys will foul out like at xavier we'll then get smoked. But until overtime happens dont worry about it.... Anyway that one simple tweak raised our post duquesne results by about 2 bpi points and I believe it to be justifiable- and to be fair I adjusted Louisville's OT loss vs notre dame number upward as well.

[DD3]

Somebody's been busy.

J/k. Nice work!

[The Wiz]

Interesting work. Nice analysis

[BACKHANDtheRICAN]

I was interested in quantifying SLU's 2nd half results. The 2nd half team is a seemingly far different entity from the first half team, through no fault of their own (Majerus, injuries). My belief is that SLU put the first half season behind them and perhaps even used it as a motivation device to become the elite team they were meant to be all along. I had requested of the wiz to see if he could isolate second half SLU numbers to give us a clearer picture of this developement. Unfortunately, it would require too much tweaking of the program he uses to provide these stats. I decided to see what I could produce on my own.

Disclaimer- understand that I am not anywhere in the league of the wiz, nor claim to be a stats guru of any sort or intend to become one. Just doing what I can to fill in the gap for informational purposes only....

I decided to use epsn's bpi rankings. These are similar to the wiz's in the sense that they incorporate margin of victory into their calculations- it doesn't just matter that the game was won or lost, but by how much you won or lost is critically important. Are you winning by getting by with the skin of your teeth in overtime or are you destroying your opponent by 30 points in regulation? Bpi incorporates other things into their calculations as well (I don't know if they really advertise everything they put into it...) but it serves my purpose in the sense that a ranking is given for every specific game and thus whatever time frame someone desires can be calculated.

SLU's Bpi number at the end of the season was 80.9 ranked #19. Calculating the Bpi based on SLU results post duquesne produced a value of 93.3 which would be ranked #1 on a full season basis (Louisville is #1 with a 91.3 value). The early season Bpi based on SLU results from duquesne and earlier produced a value of 71.6 and ranking of #56 which is about where Temple is. So according to Bpi, in the first half of the season we were playing Temple quality basketball and in the second half we have been playing to a level exceeding that of Louisville (full season). I was suprised by these numbers- its readily apparent that the team has been reborn if you will this second half but that is a vast increase. Just imagine if the whole season could have been played at this level... Another interesting thing is that when comparing the pre and post duq teams against each other, bpi projects post duq SLU beats pre duq SLU by about 14 points- which actually seems about right considering....

Another interesting calculation I made was SLU's Bpi post duq in "games that mattered most". For these I included vs top 25 teams, playing for conference reg season title (lasalle), playing for conference tournament (charlotte, butler, vcu), and against New mexico state. SLU is 8-0 in those games with a Bpi value of about 96.8. This is very good to see as it says that SLU is rising to the occasion in the biggest games- bodes well going forward.... That 96.8 value against Louisville's full season value of 91.3 would give you a projection of SLU by about 5 points. While this is a good result to see- if we are going to make adjustments for SLU we need to be fair to Louisville as well. Louisville has been playing lights out starting around the same time frame that our surge began. Louisville had a 3 game losing streak mid season but since then, they like us have a single overtime loss on the road. From the Pitt game forward, Louisville's Bpi value is 95.7. That then gives a result of SLU by 1 if we play at ultra high "games that matter" level, Louisville by about 2 if we play at average second half standards. To find that SLU may be realistically considered to be about even with Louisville is very encouraging... When just eyeballing all the other top seeds, no other team has Bpi values anywhere close to where SLU and Louisville's are (2nd half). So if SLU does get there that game becomes key as whoever wins will be favored by at least a few points in all games to follow... (granted we are viewing SLU in the most favorable light possible, disregarding the first half results, but given the circumstances I dont think it's unrealistic to do so....).

Concerning the SLU vs. Oregon matchup, Oregon has a full season bpi value of 76.9 ranked #33. Bpi has a method for adjusting when teams are missing a top player so I'm disregarding the 5-4 results when they were missing Artis. This adjusts their Bpi up to 79.5 which is ranked #23. The post duq SLU bpi value of 93.3 vs Oregon at 79.5 gives you a projection of SLU by 10 points. Using the ultra high SLU ranking for "games that matter" gives a projection of SLU by 13 points. It is important to note though that the last 3 games Oregon has played about its best of the season. For the last 3 games Oregon has a Bpi of about 97.1 If that high level Oregon team is the one that shows up its about even with our "games that matter" team. I don't know if 3 games is enough to identify a trend but its worth mentioning. I think its important for us to make Oregon aware as early as we can in the game that they are actually the #23 overall team and not the ultra high level team they are now playing at- expose the last 3 games as an anomaly rather than a trend. I'm hopeful that all of their righteous indignation at their admittedly poor seeding was discharged at the expense of oklahoma state. If they got too emotional about it they may have "spent their wad" so to speak...

I estimated SLU's Bpi value against New mexico state to be a 98. I know the game looked easy at times but that is a testament to how good SLU is playing right now. New mexico State is decent- has a bpi value of 66.4 which ranks #74. They are just below Xavier (#72) and just above Umass (#79). So it was like beating Xavier or Umass by 20 at a neutral site... One final thing- I know that Louisvile taking down NC A&T may have appeared more impressive than our win as it was by a much larger margin (31 points vs 20). NC A&T has a bpi value of 39.8 ranked 236. They had to win the play in game 2 nights earlier and it was a de facto home game for Louisville. Applying only 2 points for home court advantage (rather than a standard 3 or 4), and not considering the effect of the extra game, Louisville's bpi for that game is about 98 as well. So basically, they were about equal performances by the two teams who are currently playing at the highest level.

NOTES:

I performed one individual tweak on the Bpi calculations in regard to how they account for overtime games. Looking at the numbers, they base their calculations on what the final score was as opposed to viewing it as a tie since the two teams played to even through regulation. This is important because if you lose by 50 points in overtime Bpi calculates the result as if you had lost by 50 points during regulation- no difference. My belief is that this unnecessarily "penalizes" you for getting to overtime- you could have lost by 49 during regulation and Bpi would give you a slightly better grade for this. Had you lost by just a point during regulation you would have a night to day vastly better grade. So my view is that overtime results should not be considered in the basic calculation- it should be looked at as a "tie". This is important for SLU because the 11 point overtime loss against Xavier is viewed as an 11 point loss as opposed to a tie. The Bpi value for that game considering it to be is 11 point loss is 43.6 which is the the grade just below CS northridge. Considering it a tie gives a value of 75.2 which is the season grade between butler and stanford. In other words, my adjustment views it that we played a butler quality game that night (was just a slightly off night, which fits with what happened...) instead of saying we played a CS northridge quality game (ouch...). Now if you wish to argue that the overtime results should still be counted then I say worst case the basic numbers should include an asterisk- that they are only based on results through regulation. Now if the game were to go to overtime, ya maybe if 4 of our guys will foul out like at xavier we'll then get smoked. But until overtime happens dont worry about it.... Anyway that one simple tweak raised our post duquesne results by about 2 bpi points and I believe it to be justifiable- and to be fair I adjusted Louisville's OT loss vs notre dame number upward as well.

The Wiz is pissed. You just took his job.

[GW]

The Wiz is pissed. You just took his job.

I just couldn't handle the wiz's underestimated spreads any more. Rather than using the eastsidejoe method of just adding 9 points to the wiz's spread, I thought I'd try to do something more accurate. SLU could have played a preschool squad and the Wiz would say "SLU by 1" and then when they win by 1000 points would just shrug it off- "I told you they would win, what are you complaining about...." (j/k wiz)

[bomble]

The Wiz is pissed. You just took his job.

[bk18]

The real line is 0 at tip off. That's what matters.

[billikenbill]

Wonder what the bpi analysis would be for Georgetown and Florida Gulf Coast prior to last night's game.

[Taj79]

Just win baby.

[brianstl]

I think the Hoyas had the highest BPI sinc,e Jan. 20.

[billikenfan05]

BPI is garbage

[GW]

I think the Hoyas had the highest BPI sinc,e Jan. 20.

Good catch- apparently my unscientific "eyeballing it" method missed that... I guess the important part of that statement is "had"- after yesterday's "performance" against FGCU which probably would recieve about a 26.8 bpi rating, that drops their post Jan 20 rating from 94.6 to 90.6. The Georgetown vs. FGCU result is interesting on many levels. Forecasting the spread using bpi prior to the game would have given Georgetown by 10 using their full season value and Georgetown by 20 given the high 94.6 post Jan 20 grade. Either way obviously a huge miss. So does that make Bpi, or any predictive method for that matter, worthless? Here is my (limited) understanding of what a spread number means. Figure that going into any game there are a very large number of potential outcomes. These outcomes when plotted will form a bell shaped curve. The given spread number is the one predicted to be at the center of the curve- as many results are expected to fall at or below the spread as at or above it. So if the spread number is "georgetown by 10" then if they played theoretically 1000 times say, then 500 results would be expected georgetown +10 and under and 500 results would be expected georgetown +10 and over. With 1000 results their will be some extreme values in the "tails" of the curve- both teams would be expected to have some results where they each won by 30+ points (georgetown would have a far greater number of these). So basically the spread number tells you 1) what is the most likely outcome, on average and 2) what your chances are of winning the game (the higher the number, the greater the chance of winning). It does not tell you what will happen, only what is most likely to happen. Given this, no matter what a spread says we will always have a chance. I'm only attempting to quantify what those chances may be, not saying what will happen.

I was of course actually quite pleased to see the Georgetown result. I'm not some undercover espn operative trying to covertly promote bpi... as long as SLU wins I could care less if it was chicken entrails that called it. All of the big east results (losing 5 or 8 games, marquette barely surviving a 14 seed, while A10 was undefeated until today) thus far bode very well for us in a potential matchup vs louisville. If big east teams in general were overrated and Georgetown can win the big east (reg season) and had a bpi number similar to Louisville and got taken out by FGCU, perhaps Louisville isn't that much of a goliath after all... (either that or maybe as 05 puts it bpi is just garbage).