John

@[email protected]
33 Followers
69 Following
487 Posts
I mostly use this account to talk about #BattleBots and related things, because there are fans on instances that block my main, and my other alts have other purposes such that the followers don't want to see that much of this stuff.
searchable
JohnMar 19, 2023
A CHALLENGER APPEARS!

That's @nova37 on the controls, lil' sis on broom and simulated-arena-wall duty.

#CombatRobotics #Antweight #FingertechViperKit not #BattleBots ™
JohnMar 18, 2023
It's ALIVE!

#CombatRobotics
Show threadJohnJan 3, 2023
Well, it's much more boring this way, but here you go: Rusty's first few strengths
Show threadJohnJan 3, 2023
Rusty's "weaknesses" are less flat.

I see MissKey is garbling these graphs. It's not for file size, and converting svg->png didn't help at all. 🤷
Maybe I cutoff earlier to make a smaller graph, or upload the .dot file itself.
JohnJan 3, 2023
Rusty's "strengths" according to the WIP v0.0.1 of my graph stuff.

If the same bot appears in this and also in the opponent's "weaknesses", that adds to Rusty's "score" (according to their weights).

#BattleBots
Show threadJohnJan 3, 2023
Rusty's "weaknesses" are less flat.
JohnJan 3, 2023
Rusty's "strengths" according to the WIP v0.0.1 of my graph stuff.

If the same bot appears in this and also in the opponent's "weaknesses", that adds to Rusty's "score" (according to their weights).

#BattleBots
JohnSep 2, 2022
The big show is almost here!

That's a seriously high-octane bracket, with some very highly competitive bots with familiar names. Y'all ready?

#BattleBots #BBChamps #BattleBotsChampions #BB #BBChampions #BattlebotsChamps
JohnAug 26, 2022
Post-mortem on the experiment

First off, thank you very much to everyone who voted, including but not limited to:
@chillanarchist01 @benni [email protected]
@schokopflaster @[email protected] @totoroot
@[email protected] @ccoutreach87

We did pretty well. I judge this on two criteria.

The first criterium, and there's a proper mathematical name for this but I forget what, goes like this:
We said Bot A has an 80% chance of beating Bot B. If Bot A wins we were 80% right and 20% wrong. If bot B wins we're only 20% right.
Then, how right are we on average (arithmetic mean). If we said it was 50/50, or if the match never occurred, I don't count it either way.

The other criterium is a bit more naive but it's more intuitive to the way people play guessing games like this.
In the above scenario we implicitly "called" the match for Bot A. If Bot A wins we get 1 right, if Bot B wins we got one wrong.
Then, what % of matches did we get right. I'm calling this being "directionally right".

We were about 62% right by the first criterium and 86% right by the second. Which are both decent, especially in light of the input from @tekeous 's comment.

The most obvious implication of the divergence between the two numbers, if it is to mean anything, is that we were underconfident and hedged too much toward 50/50. Could be. Or it could be luck.

The only match we were directionally wrong on was Switchback v. Lucky. It may be of interest to note that after these polls finished I noticed there was a related poll going on on Reddit. They got the same one wrong, and were more confident (though methodology differed):

https://old.reddit.com/r/battlebots/comments/wxst4h/battlebots_champions_episode_4_live_discussion/ilstzpp/

As for me, personally, what did I get wrong?

1) I don't recall many specifics from the regular season. It may be months for us, but it was days (and in some cases less than 24 hours) for the bots. If a bot came limping out of its last match of the regular season with mechanical problems, it was likely going to have mechanical problems going into this one (Lucky). If its armor was severely, brutally abused in the regular season, we call it Big Dill.

2) I underestimated Blacksmith once again.

3) I got confused between Bloodsport and Perfect Phoenix. Similar style bot, similar color scheme. Very different team.

Regarding the polling itself

1) Too many polls

2) There were also a number of clerical errors. I've set up a bit more automation that should help with that should we try again.

3) Waiting until a poll is completely expired before allowing it to alter probabilities, and therefore alter what questions I ask, is a bit of a problem. Also hope to mitigate this through automation and the Misskey API.

Anyhow, thanks again to everyone. Until next time!

#BattleBots #BBChamps
BattleBots CHAMPIONS Episode 4 - LIVE Discussion

This week the sub voted as follows: * Big Dill 6.5% v Bloodsport 93.5% * Blade 69.4% v Kraken 31.6% * Blacksmith 96.8% v Deadlift 3.2% *...

reddit