5 min read · Published · Updated
Read the post-game scoreboard the way a coach would. Role-aware benchmarks, what each number actually measures, and how to combine them before handing the JSON to an AI.

CHAMPION LENS
Ahri
Darius
Vi
Jinx
ThreshThe Tab screen at the end of a match is dense with numbers, and most of them get misread. KDA gets all the attention, but a coach evaluating the same game pulls from at least four independent signals before forming a judgment. This guide walks through the four that matter most for solo queue improvement — KDA, CS per minute, vision score, and damage share — and how each behaves differently across roles. Read this before pasting the JSON into ChatGPT or Claude; the AI's feedback gets sharper when your prompt mentions which numbers were anomalous and why.
KDA (Kill / Death / Assist) is the most visible stat and the most overrated when read alone. Two players with identical 8/2/12 lines could have produced totally different games — one snowballed lane and built the win, the other stat-padded after teammates broke the enemy mid. Treat KDA as a downstream signal and read it together with CS difference, vision, and roam timings to recover causality.
The single most useful KDA breakdown is death timing. As a rule of thumb, two or more deaths in roughly the first eight minutes point toward a lane-setup or matchup-knowledge problem, while deaths in objective fights after the 25-minute mark point toward macro calls or an item gap. "You died twice" is not a coaching insight; when you died is. The LoL2LLM export doesn't include death times, so check the replay or add them to your prompt yourself.
For how KDA is calculated, the benchmarks by rank and role, and why chasing a "good KDA" alone stalls your climb, see what KDA is and what a good KDA looks like.
CS counts every minion and monster you last-hit. Total CS depends on game length and is misleading; the comparable form is CS per minute (CS/min), which captures wave control, attention, and how much pressure your matchup is forcing on you.
Measured median CS/min by role in JP1 solo queue (patch 16.20, n = 640 games) (range of per-tier medians):
Top: 5.92 (Iron) to 7.38 (Master+)
Jungle (minions + camps): 5.92 (Iron) to 7.22 (Master+)
Mid: 5.37 (Iron) to 7.51 (Emerald)
ADC: 6.30 (Iron) to 8.08 (Master+)
Support (you should generally not be taking minions): 0.91 (Iron) to 1.13 (Gold)
What actually matters is the gap to your matchup. Your 6.5 CS/min versus their 8.0 means losing 1.5 CS every minute — about 22 CS over 15 minutes. When you prompt an AI, telling it "CS difference vs lane opponent at 14 minutes was −22" is far more useful than "I had 110 CS at 14."
For CS/min averages by rank and concrete ways to raise your last-hitting, see what CS per minute is and the targets to aim for.
Vision score combines wards placed, wards killed, and the time-on-map value your wards generated. It's a core support metric, but it also reveals whether non-support roles are participating in the map at all.
Measured median vision score per minute in JP1 solo queue (patch 16.20, n = 640 games) (range of per-tier medians; multiply by 30 for a 30-minute game):
Support: 2.17 (Iron) to 3.00 (Master+)
Jungle: 0.88 (Iron) to 1.07 (Platinum)
Top: 0.77 (Silver) to 0.87 (Gold)
Mid: 0.66 (Iron) to 0.81 (Master+)
ADC: 0.52 (Bronze) to 0.73 (Platinum)
An ADC or mid well below their tier's median may not be buying control wards before drake or scanning before tower dives. The data backs that up: the median ADC buys 0.0 (Iron) to 1.0 (Platinum) control wards per game. Vision is not exclusively a support job; one control ward in the river before an objective contest can flip the entire game.
For vision-score benchmarks by role and why placement and timing matter more than raw ward count, see vision score by role.
Damage share is your portion of your team's total champion damage. It's the cleanest way to evaluate carries (ADC and AP mid) because it captures the contribution that KDA can't see — sustained pressure in fights, even when you don't land the killing blow. In the LoL2LLM JSON, it's your dmg_champs (damage to champions) divided by the sum of dmg_champs for you and your four teammates.
Measured median damage share by role in JP1 solo queue (patch 16.20, n = 640 games) (range of per-tier medians; an even five-way split is 20%):
ADC: 19.2% (Iron) to 25.2% (Silver)
Mid: 21.3% (Iron) to 23.8% (Master+)
Top: 20.5% (Iron) to 23.6% (Bronze)
Jungle: 14.3% (Iron) to 20.3% (Master+)
Support: 8.8% (Master+) to 13.1% (Bronze)
An ADC sitting at 18% damage share is below every tier's ADC median above — even with a clean KDA, they may not have been providing enough carry damage. A support at 18% is far above every tier's support median and probably wasn't playing an enchanter; that's a draft signal, not necessarily a player one.
For the difference between total damage and damage to champions, and how to read DPM (damage per minute), see LoL damage explained (DPM & damage share).
Single metrics mislead. A 2/8/4 support with 90 vision score, eight control wards, and the highest damage-taken on the team is doing the job — the KDA is noise. When you export a match through LoL2LLM, the JSON includes the raw numbers behind these four signals (KDA, CS, vision score, damage to champions) plus your lane opponent's, but not role benchmarks. Ask the AI to "evaluate against role expectations" with nothing else and it will make the benchmarks up, so paste the measured medians from this guide (your role's CS/min median, for example) into the prompt first.
Found this article useful? Leave a like, and share your thoughts or questions in the comments.
KDA is (kills + assists) divided by deaths, a rough measure of fight participation and survival. Measured in JP1 solo queue (patch 16.20, n = 640 games), the all-roles median KDA ranges from 1.80 (Iron) to 2.50 (Emerald) across tiers. KDA is an outcome, not a cause, though: read it alongside CS difference, vision, and death timing to understand why it happened.
Measured in JP1 solo queue (patch 16.20, n = 640 games), median CS/min is 5.92 (Iron) to 7.38 (Master+) for top, 5.92 (Iron) to 7.22 (Master+) for jungle, 5.37 (Iron) to 7.51 (Emerald) for mid, 6.30 (Iron) to 8.08 (Master+) for ADC, and 0.91 (Iron) to 1.13 (Gold) for support (range of per-tier medians). The gap to your lane opponent matters more than the absolute number: a 1.5 CS/min gap adds up to about 22 CS over 15 minutes.
Measured in JP1 solo queue (patch 16.20, n = 640 games), median vision score per minute is 2.17 (Iron) to 3.00 (Master+) for support, 0.88 (Iron) to 1.07 (Platinum) for jungle, 0.77 (Silver) to 0.87 (Gold) for top, 0.66 (Iron) to 0.81 (Master+) for mid, and 0.52 (Bronze) to 0.73 (Platinum) for ADC (range of per-tier medians). Multiply by 30 for a 30-minute game. A carry well below their tier's median is often not buying control wards before objectives or scanning before dives.
Damage share is your portion of your team's total champion damage; an even split across five players is 20%. Measured in JP1 solo queue (patch 16.20, n = 640 games), the median is 19.2% (Iron) to 25.2% (Silver) for ADC, 21.3% (Iron) to 23.8% (Master+) for mid, 20.5% (Iron) to 23.6% (Bronze) for top, 14.3% (Iron) to 20.3% (Master+) for jungle, and 8.8% (Master+) to 13.1% (Bronze) for support (range of per-tier medians). It captures carry contribution that KDA hides.
OriannaA field-by-field walkthrough of the JSON LoL2LLM hands to ChatGPT or Claude: how the export checkboxes map to JSON keys, a real output sample, and what to watch for when you paste it.
SamiraHow KDA is calculated in League of Legends, with measured median KDA by rank and role from JP1 solo queue match data, and why chasing KDA alone will mislead your climb — plus what to look at instead.
GarenWhich numbers should a beginner actually aim for to climb in League of Legends? Measured medians for CS/min, kill participation, vision score, deaths, and damage share by rank and role from JP1 solo queue match data — and what to fix first.