Cajun Softball Adjusted RPI at #12 this week ...

[Bandwagon King]

I'm guessing Regional but no national seed. That's a shame. There is little to suggest that this team is not one of the 8 best in the country. RPI might be the BCS of the diamond sports.

Only institutions of higher learning could figure out a way to lower your ranking by constantly winning.

[LFTCajun]

Only institutions of higher learning could figure out a way to lower your ranking by constantly winning.

Brian,

Because of your presence here, we are probably the most informed and educated fanbase with regard to all things RPI anywhere. Thanks for that.

I know you don't like the RPI in its current form, so what would (in your opinion) make it more accurate? In the case of UL run-ruling FAU in all three games, might the addition of that stat alone (run-rule sweep) factor into the RPI formulae? Mercy victories, margin of victories, etc? Should some consideration be given for the "decimation" factor? The mercy rule already protects most teams from the ULs of the world from running up the score, so where's the harm in rewarding a team in our situation from beating weaker teams, and convincingly?

I realize I am preaching to the choir, but just curious of your take.

[CajunRed]

Brian...considering how well you know and understand the RPI system as it is currently set up, I'm more interested in how YOU think it should be changed. I think I can list at least ONE change you'd put in place...weighting road wins over home wins...and home losses more so than road losses...like baseball will do beginning next year.

[Cajun90]

Brian...considering how well you know and understand the RPI system as it is currently set up, I'm more interested in how YOU think it should be changed. I think I can list at least ONE change you'd put in place...weighting road wins over home wins...and home losses more so than road losses...like baseball will do beginning next year.

Reference Boyd Nation's ISR for baseball. This is definitely a starting point.

[GoneGolfin]

Brian,

Because of your presence here, we are probably the most informed and educated fanbase with regard to all things RPI anywhere. Thanks for that.

Thanks LFTCajun.

I will attempt to combine a reply to two similar posts.

As a background, you can refer to this post earlier in the month.

I know you don't like the RPI in its current form, so what would (in your opinion) make it more accurate?

There are many things you could do to make the RPI more accurate than it is at present (and less biased) ... but that is not nearly enough in my opinion. It really needs to be replaced in its entirety because the core algorithm is severely lacking. As outlined in the post above, it all comes down to the RPI having a significant flaw in how SOS is calculated. To fix this, you need to develop a new algorithm. I would argue that this algorithm must recursively descend through all opponent connections until a cycle is reached for each given connection. A cycle would be defined by reaching a team that you reached in a prior traversal. Without the cycle detection, the algorithm would loop indefinitely. If your rating system is a W/L rating system (as the RPI and other similar systems are), this would be a much accurate way of determining real SOS ... with the purpose of SOS being to place your W/L record in the proper context. RPI only takes you down to your opponents' opponents.

Of course, it is more complex than the RPI formula/algorithm. But that is more of a matter for the developer writing the software. There are other systems that I am sure would be quite acceptable. The ISR should get a serious look as it is similar in nature to the above in that it uses a similar recursion to go deep to understand SOS.

In the case of UL run-ruling FAU in all three games, might the addition of that stat alone (run-rule sweep) factor into the RPI formulae? Mercy victories, margin of victories, etc? Should some consideration be given for the "decimation" factor? The mercy rule already protects most teams from the ULs of the world from running up the score, so where's the harm in rewarding a team in our situation from beating weaker teams, and convincingly?

I would not be interested in margin of victory being any part of the equation. I think it would be bad for the sport and would encourage manipulation. Even with the mercy rule in place, it might encourage coaches to overuse their ace pitcher (as an example) in the hopes of gaining favor in the mathematical system. Besides, a mercy rule is another arbitrary constant introduced into the formula that would skew margin of victory. I think it should be about winning and losing.

Brian...considering how well you know and understand the RPI system as it is currently set up, I'm more interested in how YOU think it should be changed. I think I can list at least ONE change you'd put in place...weighting road wins over home wins...and home losses more so than road losses...like baseball will do beginning next year.

So, from reading the above, you know how I think it should be changed.

If you told me the only thing I could do is "apply lipstick to the pig", I would probably do the following ...

- I would remove the ridiculous and arbitrary bonus systems. Softball is the most egregious ... in fact it is downright folly. But baseball has its has bonus system problems as well (which will be removed next season).
- Yes, weighting home and road games differently is important. But what baseball has provided for this does not fully address the problem ... in fact it barely touches upon it. Next year, a home win will count as .7 wins and a road win will count as 1.3 wins. Aside from the fact that the ratio favors road wins too much in baseball (which is another discussion), this mechanism only accounts for home/road adjustments to WP ... which is only 25% of the formula!
- The weighting between WP, OWP, and OOWP would need to change. I would also add OOOWP ... it does not fix all of the problems, but does improve the situation. I have not arrived at the appropriate weightings (simply because I have not studied the matter) ... a regression test would be required. But I can assure you that it would not be 25/50/25.


Brian

[CajunAmos]

Thanks LFTCajun.

I will attempt to combine a reply to two similar posts.

As a background, you can refer to this post earlier in the month.


There are many things you could do to make the RPI more accurate than it is at present (and less biased) ... but that is not nearly enough in my opinion. It really needs to be replaced in its entirety because the core algorithm is severely lacking. As outlined in the post above, it all comes down to the RPI having a significant flaw in how SOS is calculated. To fix this, you need to develop a new algorithm. I would argue that this algorithm must recursively descend through all opponent connections until a cycle is reached for each given connection. A cycle would be defined by reaching a team that you reached in a prior traversal. Without the cycle detection, the algorithm would loop indefinitely. If your rating system is a W/L rating system (as the RPI and other similar systems are), this would be a much accurate way of determining real SOS ... with the purpose of SOS being to place your W/L record in the proper context. RPI only takes you down to your opponents' opponents.

Of course, it is more complex than the RPI formula/algorithm. But that is more of a matter for the developer writing the software. There are other systems that I am sure would be quite acceptable. The ISR should get a serious look as it is similar in nature to the above in that it uses a similar recursion to go deep to understand SOS.


I would not be interested in margin of victory being any part of the equation. I think it would be bad for the sport and would encourage manipulation. Even with the mercy rule in place, it might encourage coaches to overuse their ace pitcher (as an example) in the hopes of gaining favor in the mathematical system. Besides, a mercy rule is another arbitrary constant introduced into the formula that would skew margin of victory. I think it should be about winning and losing.


So, from reading the above, you know how I think it should be changed.

If you told me the only thing I could do is "apply lipstick to the pig", I would probably do the following ...

- I would remove the ridiculous and arbitrary bonus systems. Softball is the most egregious ... in fact it is downright folly. But baseball has its has bonus system problems as well (which will be removed next season).
- Yes, weighting home and road games differently is important. But what baseball has provided for this does not fully address the problem ... in fact it barely touches upon it. Next year, a home win will count as .7 wins and a road win will count as 1.3 wins. Aside from the fact that the ratio favors road wins too much in baseball (which is another discussion), this mechanism only accounts for home/road adjustments to WP ... which is only 25% of the formula!
- The weighting between WP, OWP, and OOWP would need to change. I would also add OOOWP ... it does not fix all of the problems, but does improve the situation. I have not arrived at the appropriate weightings (simply because I have not studied the matter) ... a regression test would be required. But I can assure you that it would not be 25/50/25.


Brian

What specifically are the bonuses you refer to?

[GoneGolfin]

What specifically are the bonuses you refer to?

Bonuses and Penalties against the Top 25/50/XX and Bottom 25/50/XX Base RPI teams.

Brian

[CajunAmos]

Bonuses and Penalties against the Top 25/50/XX and Bottom 25/50/XX Base RPI teams.

Brian

I've read you're earlier posts, but I don't recall seeing exactly how these work. Are they listed somewhere?

[GoneGolfin]

I've read you're earlier posts, but I don't recall seeing exactly how these work. Are they listed somewhere?

See this post from March 26th.

Brian

[CajunAmos]

See this post from March 26th.

Brian

Thanks. It mentions in that post that you would provide the actual bonuses/penalties in a future post. Did you do that?