Michigan State Championship Projections

[Ed Law]

Thanks for doing this. What is your projection cutoff points at this time?

By the way, please put teams that won district chairman's into the in group no matter how their robot performs. This may reduce the number of spots left and slightly raise the cutoff points.

[Anthony Galea]

Originally Posted by Ed Law

Thanks for doing this. What is your projection cutoff points at this time?

By the way, please put teams that won district chairman's into the in group no matter how their robot performs. This may reduce the number of spots left and slightly raise the cutoff points.

For reference, all Chairman's award winners so far (7 events remain):
1322
1025
3618
33
2604
2834
66
503
3767
2619
68
548
2137
85

[oliviat3357]

Are these in order of current ranking?

[jdunston94]

Originally Posted by oliviat3357

Are these in order of current ranking?

I believe they are. here is the Blue alliances results.
https://www.thebluealliance.com/even.../2016#rankings

[SoftwareBug2.0]

Originally Posted by Ed Law

Thanks for doing this. What is your projection cutoff points at this time?

By the way, please put teams that won district chairman's into the in group no matter how their robot performs. This may reduce the number of spots left and slightly raise the cutoff points.

My best guess for the point cutoff is 69, but the range that I estimate that it could be in is pretty wide: somewhere between 57 and 106. I need to do a better job of making this explicit in the listing since you aren't the first one to ask.

There are several methodological improvements to the script that would make sense to do (like taking Chairman's into account). I'm not planning to do any of them in time for this season though. However, it is easy to rerun the script as if there were a a few less slots at MSC. For example, there are currently four Chairman's winning teams ranked below 102, and rerunning as if there are only 98 slots gives a cutoff between 59 and 107 with a best guess of 72, which is three points higher.

[Basel A]

Originally Posted by SoftwareBug2.0

My best guess for the point cutoff is 69, but the range that I estimate that it could be in is pretty wide: somewhere between 57 and 106. I need to do a better job of making this explicit in the listing since you aren't the first one to ask.

There are several methodological improvements to the script that would make sense to do (like taking Chairman's into account). I'm not planning to do any of them in time for this season though. However, it is easy to rerun the script as if there were a a few less slots at MSC. For example, there are currently four Chairman's winning teams ranked below 102, and rerunning as if there are only 98 slots gives a cutoff between 59 and 107 with a best guess of 72, which is three points higher.

Noticed in the PNW thread that you're consistently underestimating by a couple points, Chairman's could be the cause of that.

[Orthofort]

I projected the point cutoff a number of ways. (Since my team has a vested interest in the cutoff since we have 72 district points)

1. Before the season even started, I made a prediction using last years rankings. I used the ratio of the number of teams from this year to last year, multiplied that by 102 to get about 84th, subtracted a few to fudge for chairman's, and found the district points of that rank in 2015, and got my preseason projection of 71 points being the cutoff.

2. I took the current rankings as of week 4. For any team that just played one event, I added the points they got from their first event, and ranked them. This only gave a score to 391/411 teams since the other 20 hadn't played an event. Then by taking this ratio and multiplying it by 102 we get almost nearly 97. The points of 97th place in these projected rankings is 72 points. This doesn't account for chairman's, but there's also historically a smaller proportion making it from State's from those teams that play week 5 and 6, possibly because the level of competition has already gotten better. So I guessed that these two factors might cancel out. It's worth noting that I did this same method after week 4 last year, and arrived at 65 which was just about the cutoff last year.

3. I took last years district points and using the scores from highest to lowest, made a graph with the x-axis being 1st, 2nd, 3rd, all the way to 345th (ignoring two teams that didn't play events), and the y-axis being pre-MSC points. I then did a regression for this graph with an emphasis on the upper half fitting well. It ended up fitting a 4-parameter logistic curve quite well. Thus, by using the equation y=d+(a-d)/(1+(x/c)^b) I got coefficients of a=167.13, b=0.65, c=106.18, d=-43.16. I then assumed that one would need to be in top 96, thereby accounting for chairman's. Using the figures of 345 teams in 2015, 411 teams in 2016, we again take that ratio and multiply it by 96, and get 80.58. Find y(80.58) and get 71.39. Again getting that the cutoff will be between 71 and 72 (i.e. that some teams with 71 will make it).

Given that I've used three different methods, I am now willing to say with some confidence that the cutoff will be around 72. This does not reassure me (since my team has 72 points!)

[Josh Fox]

I've taken a (admittedly very crude) crack at this, and come up with 107 points as a "worst case" at the conclusion of week 4.

Here is my spreadsheet, where I took 102 slots (?) for MSC, subtracted the 14 teams already qualified via Chairman's or what-have-you, leaving 88 spots. I then used the max amount of points any single team earned at their first comp (76), and added that to every team who hasn't played a second time, and as both event scores for the teams yet to play at all. 106 points would leave you right on the bubble depending on tie breakers. This is viewable when you use the second filter view.

Obviously every team earning 76 points at their next competition isn't at all realistic, but it's early in the morning and I should be sleeping, so a more realistic set of criteria will have to wait until later.

[Ed Law]

Originally Posted by Orthofort

I projected the point cutoff a number of ways. (Since my team has a vested interest in the cutoff since we have 72 district points)

1. Before the season even started, I made a prediction using last years rankings. I used the ratio of the number of teams from this year to last year, multiplied that by 102 to get about 84th, subtracted a few to fudge for chairman's, and found the district points of that rank in 2015, and got my preseason projection of 71 points being the cutoff.

2. I took the current rankings as of week 4. For any team that just played one event, I added the points they got from their first event, and ranked them. This only gave a score to 391/411 teams since the other 20 hadn't played an event. Then by taking this ratio and multiplying it by 102 we get almost nearly 97. The points of 97th place in these projected rankings is 72 points. This doesn't account for chairman's, but there's also historically a smaller proportion making it from State's from those teams that play week 5 and 6, possibly because the level of competition has already gotten better. So I guessed that these two factors might cancel out. It's worth noting that I did this same method after week 4 last year, and arrived at 65 which was just about the cutoff last year.

3. I took last years district points and using the scores from highest to lowest, made a graph with the x-axis being 1st, 2nd, 3rd, all the way to 345th (ignoring two teams that didn't play events), and the y-axis being pre-MSC points. I then did a regression for this graph with an emphasis on the upper half fitting well. It ended up fitting a 4-parameter logistic curve quite well. Thus, by using the equation y=d+(a-d)/(1+(x/c)^b) I got coefficients of a=167.13, b=0.65, c=106.18, d=-43.16. I then assumed that one would need to be in top 96, thereby accounting for chairman's. Using the figures of 345 teams in 2015, 411 teams in 2016, we again take that ratio and multiply it by 96, and get 80.58. Find y(80.58) and get 71.39. Again getting that the cutoff will be between 71 and 72 (i.e. that some teams with 71 will make it).

Given that I've used three different methods, I am now willing to say with some confidence that the cutoff will be around 72. This does not reassure me (since my team has 72 points!)

Great analysis. I think 72 should be very close to the cutoff like you predicted. Can you redo #2 with data from Week 5 and see what the cutoff is? You should be doing this for Michigan every year from now on.

[Coach#3536]

I love to read post predictions with half the data. Way to go

[Orthofort]

Originally Posted by Ed Law

Great analysis. I think 72 should be very close to the cutoff like you predicted. Can you redo #2 with data from Week 5 and see what the cutoff is? You should be doing this for Michigan every year from now on.

Thanks a bunch!

I repeated the analysis on the data. Now, everyone has played one or two events (except one team that I presume will not attend any) and thus we don't need to fudge the number of slots for teams that haven't played. With these projected points, the 102nd slot was the very last team with 72 points. If you account for chairman's winners, which so far there are 5 winners that are below the cutoff. There may be more come week 6, but if you assume 5 auto-qualifiers, then the cutoff is the very last team with 73 points.

This is interesting because it has gone up ever so slightly from last week. But there are other factors at play. There are more teams than usual this year that are playing their second event in week 6 than other weeks (since there were relatively fewer 3rd district slots this year, and a lot of them were are LSSU, for example). This shouldn't bias the data that much, but it reduces the external factor that 3rd district teams tend to do marginally better than teams playing their 2nd district out, because they have more experience in the year. Thus, it implies that the cutoff may be closer to the chairman's fudged cutoff of 73. Either way I think it means that there's room for variance, but it is likely that either all teams with 72 points and some with 71 make it in, or just some teams with 72 make it in (ignoring declines).

On another aside, I dug up this post from last year that presents an approximation based on past years points and only is dependent on % of teams attending. If we use 102 teams out of 410 teams for this year, and plug it into that equation, we get approximately 71.39 points which I thought was just so cool since that's what my totally different regression method got (see #3 from my post). I do think that this formula gives a bit low of a number since it was a bit low last year, and that's probably because it doesn't account for the change in district points last year. I would try to add last year's data point to that data, but I don't know where to find the data from previous years that he used.

[lynca]

Michigan state championship qualification is brutal this year.
If i'm reading the predictions correctly,
great teams like 469, 862 & 910 did not qualify for the state event

Good luck to all the teams at the week 6 events !

[nikeairmancurry]

Saw an interesting thing, Team 66 has won a chairman's award, which auto qualifies them for States, yet you have them '~in".

Just wondering if that is a typo for them.

[Arhowk]

Originally Posted by nikeairmancurry

Saw an interesting thing, Team 66 has won a chairman's award, which auto qualifies them for States, yet you have them '~in".

Just wondering if that is a typo for them.

List wasn't updated since

[Zebra_Fact_Man]

Originally Posted by nikeairmancurry

Saw an interesting thing, Team 66 has won a chairman's award, which auto qualifies them for States, yet you have them '~in".

Just wondering if that is a typo for them.

Originally Posted by Arhowk

List wasn't updated since

I don't think ANY Chairman's Awards are taken into consideration in this analysis. 1322 (curently in 220th w/ 37pts) won Chairman's and is qualified for MSC, yet the newest version posted in comment #13 has them listed as "out".