Serpentine Drafting by the Numbers

In this week’s FUN Candidly Speaking, Serpentine Drafting was a popular topic.

For the uninitiated, Serpentine Drafting is how our alliance selections are done, 1 through 8, then 8 through 1.

To put some numbers behind the madness, refer to the following:

Team #1, #2 and #3 seem to have the highest event win rate, at 9.8%, 9% and 6.75% respectively.


rank	%win
1	9.884678748
2	9.060955519
3	6.754530478
4	6.425041186
5	4.612850082
6	3.953871499
7	3.294892916
8	1.812191104
9	2.471169687
10	2.306425041
11	3.624382208
12	2.306425041
13	1.647446458
14	2.965403624
15	2.141680395
16	1.97693575
17	2.141680395
18	3.294892916
19	1.97693575
20	2.635914333
21	1.97693575
22	1.97693575
23	2.471169687
24	1.812191104
25	2.800658979
26	1.97693575
27	0.823723229
28	0.988467875
29	1.647446458
30	1.317957166
31	1.153212521
32	0.658978583
33	0.658978583
34	0.823723229
35	0.329489292
36	0.988467875
37	0.823723229
38	0.164744646
39	0.164744646
41	0.329489292
44	0.164744646
47	0.329489292
51	0.164744646
52	0.164744646

All, as Karthik has requested, here is the win% plotted against the Alliance Number (1-8)

http://i.imgur.com/hHyMYot.png


rank	%win
1	30.15075377
2	24.79061977
3	15.57788945
4	12.39530988
5	3.015075377
6	5.527638191
7	4.020100503
8	4.522613065

Awesome data! But Karthik was asking for win% vs Pick number
EX:
Alliance captain 1 - 1
Alliance 1 pick 1- 2
Alliance captain 2- 3
Alliance 2 pick 1 - 4
Alliance captain 3 - 5
Alliance 3 pick 1 - 6

Alliance 8 pick 1 - 16
Alliance 8 pick 2 - 17
Alliance 7 pick 2 - 18

Alliance 1 pick 2 - 24

EDIT: Which I realize is going to just be the same data with some duplicated points, but it makes for a good visual representation of how much the late picks in the draft are getting rewarded relative to say pick 20.

Is this data just from this year’s game? If so, how does it compare to 2016 or 2015?

All,
I misread Karthik’s request. Here it is updated with the pick sequence (i.e. A1-Cap, A1-Pick, A2-Cap, A2-Pick), etc.

http://i.imgur.com/xbRl8vt.png


pick_seq	%win
1	9.433962264
2	9.777015437
3	8.233276158
4	8.233276158
5	5.317324185
6	5.317324185
7	4.288164666
8	4.288164666
9	1.02915952
10	1.02915952
11	1.886792453
12	1.886792453
13	1.372212693
14	1.372212693
15	1.54373928
16	1.54373928
17	1.54373928
18	1.372212693
19	1.886792453
20	1.02915952
21	3.945111492
22	5.317324185
23	8.061749571
24	10.2915952

All data is from 2017, accurate as to the end of Week 4 (last night).

I will compare to 2016 later, when I’m not in class.

This is cool. I’m not sure why it’s not 1 = 2 = 24 and 15 = 16 = 17 though?

I predict it’s due to Week 0. I have verified the numbers fed in are correct, as is the calculation of pick sequence. I’ll take a further look into this anomaly tonight when I have the time and see if I can hunt down why they aren’t equal

(also sorry for flooding TBA’s servers with API calls every week <3 )

Alliance 5 Winners Club represent!

Disregard my last post. The discrepancy is due to backup teams being called.

There’s some more discussion going on in this thread, demonstrating what is causing the lesser win-rate of the #1 alliance this year.

You might have forgotten to multiply by 3. The 1 seeded team should have close to the same win % as the 1 alliance.

This is true when comparing to alliances, when comparing against other teams the % is based on that entire set.

I’m going to run some more numbers, including this modification and also including more recent events.

All, I have updated the numbers to better calculate backup teams, and to also include more recent events. This also scales the numbers properly for 3-robot alliances.

Win% vs Seed Rank
http://i.imgur.com/Gr9Plh0.png

Win% vs Alliance Number
http://i.imgur.com/qgcf88K.png

Win% vs Pick Number
http://i.imgur.com/hA54H3Q.png


Seed	Win%
1	35.09060956
2	23.723229
3	13.34431631
4	17.79242175
5	14.33278418
6	13.34431631
7	8.401976936
8	3.459637562
9	6.919275124
10	6.919275124
11	11.36738056
12	7.413509061
13	3.459637562
14	6.919275124
15	8.401976936
16	7.907742998
17	7.413509061
18	9.390444811
19	2.965403624
20	3.953871499
21	2.471169687
22	3.953871499
23	7.413509061
24	5.436573311
25	2.965403624
26	4.448105437
27	2.965403624
28	3.459637562
29	5.436573311
30	2.965403624
31	3.953871499
32	1.482701812
33	0.494233937
34	3.953871499
35	1.482701812
36	2.471169687
37	0.988467875
39	1.482701812
40	0.494233937
41	0.494233937
44	0.494233937
47	0.494233937
52	0.494233937


pick_seq	win%
1	43.93063584
2	48.55491329
3	17.34104046
4	19.65317919
5	10.98265896
6	10.40462428
7	14.45086705
8	15.60693642
9	4.046242775
10	4.624277457
11	2.312138728
12	2.312138728
13	1.156069364
14	1.156069364
15	4.046242775
16	3.468208092
17	4.046242775
18	0.578034682
19	1.734104046
20	4.624277457
21	13.87283237
22	9.248554913
23	17.34104046
24	44.50867052


alliance	win%
1	45.66473988
2	18.11175337
3	10.21194605
4	14.64354528
5	4.431599229
6	2.119460501
7	0.963391137
8	3.853564547

What is this new alliance win% calculating versus the previous one? Which one is the % of events won by each alliance rank? (also thank you for this data, I’m a data nerd).

All percentages are the percentage of times they have won the event.

The old calculation failed to handle cases where a backup team was called properly (it looked for finals matches, and then filtered out which alliance had the higher score).

The new calculations use the “Regional/District Event Winner” award as its basis.