Predicting Actual High Efficiency Goal Shooting at Worlds

[Mitchell1714]

1 out of _5_ teams will have a median H.E.G. ball count that is >50
1 out of _10_ teams will have a median H.E.G. ball count that is >100
1 out of _500_ teams will have a median H.E.G. ball count that is >200
1 out of _all FRC_ teams will have a median H.E.G. ball count that is >400

I have my estimates saying fewer robots will make 50+ or 100+ because of ball control issues. Many teams will have shooters that can shoot 3 balls a second at 70% accuracy. The problem is, they won't be able to effectively load balls from their hoppers to their shooters. Either, balls will load to slow, inconsistently or have jamming issues feeding from the hopper. I think about 10 teams will have figured out their ball management systems and had enough driver practice to score 200+ balls on average. I was originally going to say no teams would average 400+ balls a match, but then I thought of 2056. Every year since 2009 they have had simple, non-jamming machines that are just monstars at picking game pieces off the ground. That's just my 2 cents

[Ether]

Originally Posted by Joe Johnson

Consider the teams at Worlds playing in the qualifying matches. For each team you can calculate a median number of balls that they scored in the High Efficiency Goal

Hi Joe,

I don't know about others, but I did not respond because I'm still pondering what you meant by a team's median ball count.

The FRC API will almost certainly not provide a ball count per team for each match, so you can't compute the median of a team's ball counts for all qual matches a team played.

If the FRC API provides (which it most likely will) alliance ball count for each match in a qual event, then it's a simple matter to use the favored OPR model and algorithm (linear least squares) to estimate an OPR-like average of each teams per match ball count for each qual event. But what does "median" mean in that context?

So perhaps what you actually meant was a percentile distribution of those team ball count "OPR" estimates? That interpretation seems to reconcile better with the fill-in chart.

[Ether]

Originally Posted by Ether

So perhaps what you actually meant was a percentile distribution of those team ball count "OPR" estimates?

Something like this, except a single percentile curve of the vector of team "OPR estimated" ball count instead of alliance scores.

[Joe Johnson]

Originally Posted by Ether

Hi Joe,

I don't know about others, but I did not respond because I'm still pondering what you meant by a team's median ball count.

<snip>

Here is what I mean.

Suppose you have a way to count the balls that a particular robot actually shot, entered into the top of the HEG and was actually counted by the scoring system; Let's call this number God's HEG Count for each robot for each qualifying match, which she keeps in God's STEAMWORKS Ledger (of course).

The Devil (being a detail guy of course) then hacks into The Cloud (St. Peter is famously lax on security) and gets a copy of God's STEAMWORKS Ledger. He imports the whole thing into Excel [ASIDE] I know what you are thinking, he's in Hell, why isn't he trying to do stats using Google Sheets? Funny story, it seems that they have a special ring there just for that purpose but... it is controlled access... and while the Devil could figure out a way in, he misplaced his ID while picnicing last weekend near Styx... and facilities in Hell just went to 2 factor authentication,... And MS Office came loaded on his work PC... Just like everyone else, it's just simpler for the Devil to use a MS product than find an alternative [/ASIDE] where he sorts God's HEG Count for each team from high to low then goes half way down the list where... VIOLA! ... He finds each team's Median HEG Ball Count (of course, if there are an even number of qualifying matches, he averages the middle two rows -- I told you the Devil is a detail guy...)

More on what I WANT from a stats person in another message...

Stay tuned...


Dr. Joe J.

P.S. This is all tongue in cheek. Nobody get offended please.
P.P.S. And if you are offended, I am sorry. JJ

[efoote868]

Originally Posted by Joe Johnson

Here is what I mean.

Why does the stat have to be median? Would you expect a team's HEG Count distribution to be vastly skewed?

[Ether]

Originally Posted by Joe Johnson quote#1

IF some clever FIRST stats person can figure out a way to get a reasonably good estimate of this number from the publicly available numbers that FIRST publishes

Originally Posted by Joe Johnson quote#2

Here is what I mean. Suppose you have a way, [ed: for each match], to count the balls that a particular robot actually shot, entered into the top of the HEG and was actually counted by the scoring system [ed: for that particular robot] [ed: and find from that] each team's Median HEG Ball Count

I think what I described (i.e. a vector of team "component OPR style estimates") is the closet thing to what you want that can be extracted from the publicly available numbers that FIRST publishes.

But I will quietly resume my lurking, awaiting a cleverer stats person to figure a way extract what you have described in the second quote from the publicly available numbers that FIRST publishes.

[Joe Johnson]

Originally Posted by efoote868

Why does the stat have to be median? Would you expect a team's HEG Count distribution to be vastly skewed?

Don't you expect a lot of zeroes in there because for one reason or another the team was asked to do something else that round (run gears for example)?

I felt that median was a better number for my thought experiment because it gives a higher number than mean with all perhaps 1/3rd of the scores being zeroes due to nobody else in the match that can effectively run gears.

That said, I suppose that Mean is probably about as good as a metric and would be a lot easier to estimate.

Dr. Joe J.

[Sperkowsky]

1 out of 5/5 teams will have a median/max H.E.G. ball count that is >50
1 out of 8/5 teams will have a median/max H.E.G. ball count that is >100
1 out of 50/25 teams will have a median/max H.E.G. ball count that is >200
1 out of 1500/750 teams will have a median/max H.E.G. ball count that is >400

[Lil' Lavery]

Among the subset of teams playing at Champs

1. 1 out of 4/3 teams will have a median/max H.E.G. ball count that is >50
2. 1 out of 10/4 teams will have a median/max H.E.G. ball count that is >100
3. 1 out of 50/30 teams will have a median/max H.E.G. ball count that is >200
4. 1 out of None/100 teams will have a median/max H.E.G. ball count that is >400

[efoote868]

Originally Posted by Joe Johnson

Don't you expect a lot of zeroes in there because for one reason or another the team was asked to do something else that round (run gears for example)?

I felt that median was a better number for my thought experiment because it gives a higher number than mean with all perhaps 1/3rd of the scores being zeroes due to nobody else in the match that can effectively run gears.

That said, I suppose that Mean is probably about as good as a metric and would be a lot easier to estimate.

Dr. Joe J.

I'm not sure what to expect, but I do know the game isn't played in a vacuum. Some matches will feature dedicated gear runners or low goal dumpers (robots that can't shoot in the HEG), and that would inflate a team's HEG count.

If there are several full field webcasts (with archives), it might be possible to try and reconcile the numbers for a few teams. I'd be interested in doing that.

[hionwind]

Originally Posted by WesleyK

1. 1 out of 4 teams will have a median H.E.G. ball count that is >50
2. 1 out of 118 teams will have a median H.E.G. ball count that is >100
3. 1 out of 330 teams will have a median H.E.G. ball count that is >200
4. 1 out of 1678 teams will have a median H.E.G. ball count that is >400

Seems to be some west coast bias here.

[Joe Johnson]

I am curious as to how folks can have a Median/Max where the first number is larger than the second number.

If for example you have an answer of the form:

1 out of A/B teams will have a median/max H.E.G. ball count that is >N

It seems to me that A < B.

To say otherwise says that there will be more teams with a 50th percentile H.E.G. ball count above N than there are teams who's 100th percentile H.E.G. ball count above N.

Said another way, how can there be more teams with their middle score over N than had their best score over N.

Have I mistated the question somehow that is confusing people?

Dr. Joe J.

[crazybobcat]

On the contrary, wouldnt A need to be greater than B? Since they are both ratios, and 1/4 < 1/2. It seemed like everyone did it correctly.
With that stated, here are my guesses
1 out of _4/2_ teams will have a median/max H.E.G. ball count that is >50
1 out of _10/5_ teams will have a median/max H.E.G. ball count that is >100
1 out of _75/20_ teams will have a median/max H.E.G. ball count that is >200
1 out of _infinity/100_ teams will have a median/max H.E.G. ball count that is >400

[Tuba4]

Originally Posted by Ginger Power

The reason I think so few teams will score high numbers of Fuel at champs is because I believe cycling gears will prove to be the priority well into champs qualifications. I doubt alliances will quickly and consistently cycle 12 gears, and leave enough time to do a significant amount of Fuel scoring. Many alliances will realize they have no hope of getting to 12 Gears, and they'll start scoring Fuel after their 6th Gear. Also, many teams will understand there is a very small benefit to scoring more than 40kPa during quals so they'll stop at that point and focus on gear cycling/defense/climbing depending on how much time is left in the match.

Alliances only need to cycle 9 gears. 3 are pre-populated. However gears 6,7 and 8 by themselves get you no points. You have to get to gear 9 to get the final points. I suspect the majority of teams will stop at gear 5.

Would this change your thinking on the number of balls scored?

[Ginger Power]

Originally Posted by Tuba4

Alliances only need to cycle 9 gears. 3 are pre-populated. However gears 6,7 and 8 by themselves get you no points. You have to get to gear 9 to get the final points. I suspect the majority of teams will stop at gear 5.

Would this change your thinking on the number of balls scored?

The three pre-populated gears are in addition to the 12 gears that need to be cycled by alliances. In total, there will be 15 gears in the Airship to get all 4 Rotors spinning, but only 12 of them need to be brought there by robots. I would guess those pre-populated gears were placed there in case the gears are less of a challenge than the GDC expected, so they can increase the difficulty before champs.