Helldock

Opensourcing my scrim tracker for valorant for t2/t3 teams so u can know whats actually going wrong and fix the mixtake one by one,google sheets link in replies

@Mazzii2x Proud of u my sonnn

LFT...

i was building my portfolio as an analyst for Titan esports,with only vlr data (Emailed grid for more data thoo) was thinking i would share this here too,mada an kind of opps dosier of every team in china,attached full work on reply

ya ur right those features alone are basic. thats v1 on purpose — score+econ+side+round catches like 70% of round outcome signal and gives clean baseline predictions. real value isnt the model being fancy, its the gap between what the model expects and what actually happened. thats where the coachable rounds live v2 will add FB/FD, util spent, map. but more features ≠ better queue. the basic version already finds rounds humans miss

@helldock yeah i mean i wanted to understand what the scores are based on. seems too basic to take it into consideration if its only the things you are talking about

Won 13-8 In scrims. felt good. didnt watch vod but one round we shouldnt have won. full buy, everything went wrong, someone clutched a 1v3 and saved it next scrim no clutch. lost 8-13. no clue why winning hides the bad rounds. you gotta watch the wins you shouldnt have won built a model for @SailOnPast that knows when u shouldve lost — or shouldve won. learns ur team round by round and flags the rounds that actually mattered, the ones hiding inside ur wins and losses Still the model need more data tuning

two scores in the screenshot bro the big yellow one (75, 60, 58) is the queue score — how worth watching the round is. higher = more u need to vod it the WP % (0%, 56%) is what the model thought ur chance of winning that round was before it played. low % + u won = surprise round = goes to top of queue the model uses score diff, side, econ, round num, pistol — trains on ur team specifically so it learns how u play

@helldock how does it measure the score?

@NobleVALORANT Yep that’s true on higher end of t2 and t1 but when an t3 and t2 with friends who doesn’t have much time to vod it will be useful I guess

@helldock This is cool and all but realistically you would skim each round anyways and then based on things you see quickly you further analyze.

last time i dropped a scrim tracker yall said it was too much fair built a lighter one. round-by-round logging, auto dashboard, no 20 sheets free. link in replies rt appreciated 🙏

@moeratmo yeah i get it,so eng is not my first language so i throw what i wanna say at twitter in ai it corrects the grammar and stuff

@helldock AI slop post

VAL T2/T3 teams scout trialists on 3 stats: KDA, FB, FD. That's why your roster looks stacked on paper and implodes at LAN. Built a 9-metric framework so you stop picking on vibes ↓

Built this for my own team's trials at Helldock. Sharing the sheet free — VAL scene needs better trial infrastructure. Sheet: docs.google.com/spreadsheets/d… Run it on your next trials. Reply with what it caught . RT if your captain or coach needs this.🫰

Quick reads I check every trial: 🚩 High K/D + low KAST = stat-padder 🚩 High FB + high FD = reckless, will tilt 🚩 Low util DMG on initiator = kit-wasting 🚩 Comms ≤2/5 = will collapse team comms ✅ KAST ≥70% = round-impact player ✅ Coachability 5/5 = skills compound from here

All 9 stats roll into a verdict — weighted by role. Combat / Impact / Process composite: • Duelist: heavy Combat + Impact • Sentinel: balanced • IGL: heavy Process Output: SIGN / BENCH TRIAL / PASS Numbers, not vibes.

Stats #6–9: The non-stat signals. Most trials are won HERE, not on the scoreboard: • Comms (info + IGL ability) • Mental / tilt resistance • Adaptability • Coachability ← single highest-leverage trait 1.3 K/D player who tilts at 0–3 = liability. 0.95 K/D player who calls clean = foundation.

Stat #5: Util damage per map This is where 90% of teams miss the initiator/controller eval. Sova with 200+ util DMG/map → winning rounds invisibly Sova with <80/map → wasting his kit, give the slot to someone else Same logic for KAY/O, Skye, Brim, Viper.

Stat #4: Clutch rate (1vX won / attempted) Pressure exposes everything: – aim breaks – panic util – bad timings A trialist who's 4/9 in clutches just showed you his floor. 30%+ = composure under pressure. Rare and expensive.

Stat #3: Multikill rounds (count of 2K/3K/4K/5K rounds) Two 3Ks swing more games than six 1Ks. Always. Multikill density = round-carrying capacity. Flat 1Ks = scoreboard padding. Track the rate per map, not just total kills.

Stat #2: FB minus FD (net entry impact) Raw FB count rewards stat-pads on retakes. FB-FD shows if he's winning the duels he picks. Duelist: +3 or better per map Sentinel: near 0 (not entering = good) Initiator: -1 to +1 range

Stat #1: KAST% % of rounds the player got a Kill, Assist, Survived, or got Traded. The single most underrated stat in trials. • T1 pros: 70–75% • T2/T3 sweet spot: 65%+ • <60% on a trialist = red flag no matter how he frags

A duelist with 1.3 K/D and 58% KAST isn't a star. He's a stat-padder farming kills in lost rounds. A controller with 0.9 K/D and 76% KAST is winning rounds you don't see. KDA tells you who got kills. It doesn't tell you who won the round