What TankCoach does

The project is moving fast — features are added almost every week, and something on this page may already look different. As of 14.11.2026. News — on Telegram.

In short

TankCoach is a coach for World of Tanks (Wargaming) and Mir Tankov (Lesta), free during the test period: a mod quietly records your battle, and a minute later the website shows a detailed breakdown — where each shell went and why, when and by whom you were spotted, what your connection did, how much stronger your team was and what is worth training.

How it differs from usual stats sites:

Site: tankcoach.net. World of Tanks reports live on app.tankcoach.net, Mir Tankov reports on ru.tankcoach.net. Almost everything is the same for both games; differences are marked (the main one: a simplified “Team strength” on Lesta, see below).

In short: how TankCoach differs from stat sites

Stat sites — such as tomato.gg, or XVM and similar mods in game — answer “how good am I overall”: WN8 and other ratings, per-tank stats, marks of excellence, comparison with other players. Battle results are enough for that: damage, spotting, win or loss. According to its own site, the tomato.gg mod uploads the results of every battle and your garage loadouts.

TankCoach answers a different question: “what happened in this battle, and why”. The mod records the battle itself — movement, aiming, every shot, your connection second by second — and the server breaks it down shot by shot:

Stat sites (tomato.gg etc.)TankCoach
Inputbattle resultsa recording of the whole battle: positions, aim, shots, hits on you, connection
Main questionwhat is my level, which tanks work for mewhy I missed, why I got penned, is the game rigging this battle
Where the shell wentnoevery shot in 3D on the map terrain, with your and the server reticle
Connectionnoping, packet loss and lag spikes per second, lag right before a shot
Armornotarget armor against your shell, your armor against every incoming hit
Ratings, leaderboards, marksyes, their strengthno — not what we do

Not “instead”, but “alongside”: stats show what works for you, TankCoach shows why. Comparison based on tomato.gg’s public descriptions as of 10 Nov 2026; they may have features we don’t know about.

Getting started

Setup takes a couple of minutes; no application and no game password are needed.

  1. Download the kit at tankcoach.net/download — one for World of Tanks, one for Mir Tankov. The archive holds the mod as a single file (.wotmod or .mtmod) and a .bat installer. The kit is signed: the installer checks the signature, finds the game folder, backs up the previous version and leaves other mods alone.
  2. Play a battle. The mod records it quietly into Netprobe_mod_result in the game folder.
  3. In the garage press Ctrl+Shift+R (or Mods menu → TankCoach → “Open → Build a battle”). The mod page opens with the list of recorded battles.
  4. Press “Build” next to a battle. The first time a consent window explains what is sent and where. After “Agree” a login is created automatically from the nickname in that battle.
  5. A few seconds later the report opens in the browser. It also appears in “My battles” on the site, which works from any device without the game.

Login and alt accounts. TankCoach has its own login: nickname and key. The mod sees which account you are playing and uses the matching login. “My battles” from the in-game menu opens the site already logged in. To log in from a phone you can set your own password at app.tankcoach.net/key.

What leaves your PC:

When What Game
On “Build” the files of one battle for analysis both
At battle start team rosters: account IDs and tanks, no nicknames — for “Team strength” World of Tanks
During and after the battle public player stats from tanki.su, no nicknames — for “Team strength” Mir Tankov
On a mod error an error report (can be turned off) both
Once a day in the garage check for a new version both

The game password is never needed and never requested. Details: “What data we collect” (tankcoach.net/privacy).

In the game: what the mod does

In battle the mod only records: it never shows enemies, hidden data or hints. Load on the game is about 1.5 % of frame time (measurements: tankcoach.net/nagruzka).

What it records (only what your game client already knows): movement and HP of vehicles the client saw; your reticle and the server reticle; every shot and hit, including the point and angle on the armour; spotting and visibility rays; the event feed; per-second connection; battle results. On Wargaming maps with “random events” (Himmelsdorf, Prokhorovka, Ruinberg, Old Harbor) it also records the event zones; players breaking trees and houses is not recorded.

In-battle indicators — only about your connection, each can be turned off:

Mods menu in the garage → TankCoach. Indicator checkboxes (the section switch hides everything visible at once) and the “Open” list: “My battles on the site” (browser, already logged in) or “Build a battle” (the mod page). Without the menu libraries the same checkboxes are on the mod page.

Mod page (Ctrl+Shift+R in the garage). List of recorded battles with “Build” buttons, indicator settings and a “battle found / battle in N s” bar so you do not miss the start while in the browser. Once a day the mod tells you in the garage if a new kit is out.

The analysis is built on our server — nothing is analysed on your PC.

“My battles” on the site

“My battles” lists all your analysed battles on one page, from PC or phone, without the game running. Address: app.tankcoach.net/battles (World of Tanks) and ru.tankcoach.net/battles (Mir Tankov). 50 battles per page, newest first.

My battles

Fig. 1. “My battles”: filters, “what the battle is remembered for” tags, statuses and buttons.

Header. “My battles” · “Connection history” (next section) · on the right the nickname and “log out” · page language RU / EN.

Filters. Period (all / today / 7 days / 30 days or custom dates), tank and map — the lists show the battle count next to each name.

Banners. Yellow — test period: analysis may be delayed, the server keeps the last 20 reports, older ones can be downloaded (battle statistics are kept forever). Green — the account has premium: reports are kept without limit.

Columns. Date · Map · Tank · Mode · What the battle is remembered for · Status · buttons.

“What the battle is remembered for” tags. Up to three most important per battle, the rest under “+N” (hover to see all). Green — good, red — bad, yellow — worth watching. Hover a tag to see its numbers (e.g. “team strength by WN8: ours 1444, theirs 1165”).

Tag Colour When it appears
Pulled it off green win with pre-battle chance below 35 %
Threw the lead red loss with pre-battle chance above 65 %
Dealt ×N HP green damage at least 3× your own HP
Carried: N frags green 3 frags or more
Our edge +N % green our team stronger by WN8 by 20 % or more
Out-balanced −N % red our team weaker by WN8 by 20 % or more
Lag N % red lag in 5 % of battle seconds or more
Lag on the shot ×N red at least 2 shots with lag right before them
Flicks ×N red 3+ shots during a swing faster than 15°/s
Unaimed ×N red (yellow for drums and autocannons) at least 5 unaimed shots and at least a third of all
Tanked Nk green armour blocked 1000 damage or more
Sniper N/N green penetrated 90 %+ of 6+ hits
Non-pens ×N red hit but did not penetrate 4+ times
Died at M:SS red destroyed in the first 3 minutes

Artillery is not rated on penetration and aim time. Team strength tags appear only if “Team strength” was computed for the battle.

Statuses. “ready” — the report can be opened; “queued” (with “N ahead” or “next”) and “building” — the server is still working; “failed” — the report did not build (the server retries after a fix); “old version” — built by an earlier builder and lacks new features (only the server can rebuild it; it rebuilds recent battles itself); “report deleted” — removed by the test-period limit, the battle row and statistics remain.

Buttons.

Report views are counted anonymously: time on screen and opened sections, no cookies or IP.

“History”: connection, balance, “What to train” and economy

The second page of the site looks at all battles at once: how your connection changed, what decided a day, and where you lose the most. Six tabs at the top and shared filters: tank, period and source (“TankCoach server” or older battles imported from netprobe 6.x on the PC).

Battles

A connection table for every battle. The line above gives the period totals: battles, median ping, “server silent” share, minutes in battle.

History: battles

Fig. 2. “Battles” tab.

By day

The same numbers summed per day — handy when talking to your ISP. The key figure is “server silent” (weighted by battle length), then “seconds with gaps”, the number of battles with each rating (green / yellow / red figure) and the day’s team strength.

By day

Fig. 3. “By day” tab.

Shell and server circle

A fairness check of dispersion on your own shots: where shells landed relative to the circle the server actually used for the shot. The pink circle is the server circle (1R), dots are shells (green inside, red outside); on the right the distribution of distance from the centre in fractions of the radius; below, the list of shells outside the circle with the reason (lag, delayed confirmation, artillery). In the author’s archive 96 % of 2932 shots landed inside, median 0.34R.

Shell and server circle

Fig. 4. “Shell and server circle” tab.

Balance: why the day went this way

The answer to “is the game rigging my battles”. Pick a day or the whole period and the page explains what decided the result and splits the deviation of the win rate from expectation into parts (in percentage points):

Below are period cards (how you played, team strength, wins, position in the list, damage · assistance · spotting), a “what the wins were made of” chart, a per-day chart (team strength and your form of the day with confidence whiskers) and a table of days. Battles with bots (fewer than 12 live players per team) are excluded.

Balance

Fig. 5. “Balance”: period breakdown and cards.

Balance by day

Fig. 6. “Balance”: team strength and form by day.

What to train

The main tab for improving. The chain: measured → found where you lose the most damage → exercise → check in the next battles. You are compared with yourself: the last 50 battles against your whole archive. Lag is subtracted: misses and deaths during bad connection are not counted as your mistakes. Random battles only.

What to train

Fig. 7. “What to train”: top leak, exercise and the 10-battle goal.

“Where you lose the most”. The server checks a set of typical mistakes and shows the two most expensive in damage per battle. Each has a cost “≈ N damage per battle” with a 95 % confidence interval (advice is shown only if the whole interval is above zero), an explanation and three examples from your battles with an “open the moment in the report” link.

What was found Exercise
Early deaths under fire from several enemies Celivalnya, “Target switching” (peek — shoot — back); Vyzhivalnya (soon)
Own aiming misses: off target and no lead Celivalnya, “Moving target”
Shots without full aim Celivalnya, “As in game”
Shell into terrain in front of your tank “Shot mechanics” (barrel below the line of sight)
Shooting where you can’t pen Celivalnya, “Edge”
Side exposed where your front would hold Tankovalnya (soon), “Angle from the side” duel

“Goal for 10 battles”. For example “no more than 1 bad battle in 10”: a “Start goal” button, ten cells filled by the next battles. A fair “helped / didn’t” verdict comes after 30 battles. The “Progress” chart shows the share of bad battles over a rolling 20-battle window against your usual level. With 20+ battles of one class in the window, separate “class advice” appears.

“Aiming before the shot”. Per tank (chips with shot counts): the last 30 battles, standing shots at an identified target. A shot is “aimed” if the server circle is at most 1.05× the fully aimed one; “partly aimed” up to 3×; “snap shot” wider. Bars show the hit rate by distance (under 100 m, 100–200, 200–300, 300+ m). If the difference is clear, advice “Aim longer — +N points of hits” appears. For drums and autocannons being unaimed inside a burst is normal, so the first shot of a burst is counted separately with “Aim your first shot” advice. Shots on the move are shown on a separate line; artillery is not rated.

Aiming before the shot

Fig. 8. “Aiming before the shot”: the second leak and the per-tank block.

Collapsed below: “Other leaks” (all checked mistakes with their cost), “Battle indicators by class” and “How it is calculated”. The page does not promise “+N % wins” — it shows where you lose.

Economy: where you earn and where you lose credits

Credits over all your battles — exact sums from each battle’s results, no estimates. At the top, a period breakdown: what you earned net, per battle and per hour, which tank farms best and which loses credits, what you missed without premium, what missions, the pet and reserves gave, where the costs went. Its own filters: tank, battle mode, period.

Economy: period breakdown

Fig. 9. “Economy”: period breakdown and totals (the screenshots show the Russian version).

Economy: where credits come from, modes

Fig. 10. “Economy”: where credits came from and where farming pays more.

Economy: tanks

Fig. 11. “Economy”: what you earn on.

The tab on the author’s real battles →

The battle report

A report is one page in the browser (best on a computer). On first open a short four-step tour shows the essentials; bring it back with the ☰ menu → “Show hints”. You can look at examples without your own battle (nicknames replaced): LT, Strv m/40L, MT, T-28, Mir Tankov, SPG, FV3805.

Battle report

Fig. 12. Report: timeline, map player, shot and tanking panels, event feed.

On screen

Sounds in the report

The player gives the battle sound: shots, hits, the spotting lamp, “Penetration!”, “Ricochet”, “armour held”. Turn it on and off with the sound button next to the map. How the report sounds depends on where you watch it:

During a battle the report does not touch game sounds: the mod provides them only in the garage.

Shot “Breakdown ▸”

The “Breakdown” button opens the shot in 3D on the real map relief: shell flight, your gun and the target, obstacles; “▶ 2 s before the shot and flight” replays the motion of your reticle and the target. Views: “from the approach line”, “side”, “top”, “from target to you”, “to the burst / obstacle”. Below are the reticle circles (yours, the server one at the click and when the server processed the shot), a “how the shell flew and the target moved” animation, “what happened” in words, gold vs regular ammo, and “move the aim” — what another aim point would have given. For artillery — the dispersion ellipse on the ground and the dead zone.

Lines in the 3D breakdown. The light-green line with dots shows how you moved your reticle during the 2 seconds before the shot (a dot every 0.5 s). It shows at a glance whether you tracked the target smoothly, jumped the reticle around, or were still adjusting when you clicked: a short line with dots bunched together — the reticle sat on the target; a long stretched line — a shot while the reticle was still moving. Below the picture the same in numbers: how many metres the reticle moved and how the aiming circle changed. Next to it — the target’s path (dots every 0.5 s: yellow — the click, red — the impact), the shell arc, the hit point and the target’s armour coloured for your shell. What the mod records around a shot (reticle movement for 3 s before it) the server also uses for advice in “What to train”.

Shot breakdown in 3D

Fig. 13. Shot “Breakdown” on the map relief.

Shot verdicts: flick (fast swing), not aimed, too little or too much lead, over/undershoot, lag before the shot, into cover (bump, rock, house), disputed, and plain dispersion.

Battle summary and the blocks below the map

Battle summary — three numbers (dealt / received, seconds on a line of fire, “cost of 1000 damage” — how much of your HP you gave per 1000 of theirs) and behaviour cards with a figure and an explanation. The stripe colour: red — a mistake, yellow — worth attention, green — good. Cards appear only when there is something to say:

Further down: Team battle (front line and “vehicles left” over time), Battle flow and Peeks in order.

Battle summary

Fig. 14. “Battle summary”: numbers and behaviour cards.

Credits this battle

The coin button to the right of the map. Net for the battle, earned and costs (repairs, ammo, consumables) — exact sums from the battle results; on World of Tanks the earnings are split by our model into taking part, victory, damage, damage on your spotting and spotting. The shot table: every shot, its price and how many credits its damage brought; clicking a row jumps to the shot.

Credits this battle

Fig. 15. “Credits this battle” in the report.

“Why misses” and “Connection” panels

Why misses sorts all misses of the battle into groups: own mistake (did not wait for aim, no lead), into cover, desync with the server, dispersion, disputed, no data. Click a line to see the shot diagram and jump to it in the report.

Why misses

Fig. 16. “Why misses” panel.

Connection in this battle — average and 95th-percentile ping, jitter, seconds with no server data, number and length of gaps, overall rating; “Shooting and connection” — which shots coincided with lag; “Tank control” — how many movement commands (drive, turn) you gave during seconds with no data from the server (“blind”), and whether there were “hull jerks”. A hull jerk is when your tank on screen suddenly snaps back or turns by itself. It happens because your client moves the tank immediately without waiting for the server, while the real position is computed by the server. If lag made them disagree, the server sends its position and the client abruptly moves the tank there. Players call it “the tank got yanked” or “teleported back”: you drove out from behind a rock and the server put you back under fire. The report counts a correction of more than a few degrees beyond what the tank could turn on its own as a jerk, and shows the largest one.

Connection

Fig. 17. “Connection” panel.

Team strength: World of Tanks and Mir Tankov

“Team strength” shows how much stronger or weaker your team was by WN8 and what the win chance was before the battle started. On World of Tanks it is complete; on Mir Tankov it is simplified for now.

Team strength

Fig. 18. “Team strength” panel (World of Tanks).

What is in the panel. A one-line verdict (“On paper your team is slightly stronger: +139 WN8 per tank. Chance ≈ 69 %”) and a red-to-green chance bar; an allies / enemies / Δ table (average WN8 on the battle tank, overall WN8, median, players with stats, damage, frags, survived); a “win rate and team strength” chart from your battles (this battle’s bar outlined); “Rosters” — every player with overall and per-tank WN8, battles on the tank, damage and frags (sort “by strength” / “by damage”, the boxed number is the platoon). The second tab “In battle: realisation and flow” shows how players actually performed against expectation.

WN8 colours: red <300 · 300 · orange 450 · yellow 650 · 900 · green 1200 · 1600 · blue 2000 · 2450 · purple 2900+. “?” and “<20” — few battles on the tank, overall WN8 used; “≈” — a new tank, expected values taken from similar tanks.

World of Tanks Mir Tankov (since 14.11.2026)
Source of player stats official Wargaming API, queried by the server public player pages on tanki.su, fetched by the mod and passed to the server (Lesta has not issued an API key yet)
WN8 inputs damage, frags, spotting, base defence, wins — all real damage, frags, wins — real; spotting and base defence taken as the tank average — WN8 marked “≈”
Anonymous players from battle results from battle results (the mod queries them in the garage)
Older battles filled in automatically only battles after installing mod version 2026.11.13
Strength tags and “Balance” available will appear as battles accumulate; numbers are approximate

On Lesta the panel shows a warning about this and “tanki.su” as the source. The server measures how far “≈” is from full WN8 on World of Tanks data, where spotting is known. Once Lesta issues an API key, the server switches to full data automatically.

How to read the chance. It is computed from the “WN8 on the tank” difference; the model is personal (your battles), general (all TankCoach players’ battles) or a default curve. Team strength explains a single battle only partly: the model does not know the map, vehicle roles or the form of the day. The panel explains this under “How the chance is calculated and how much to trust it”.

Browser trainers and how they connect to the analysis

The goal is not only to show a mistake but to let you train it and then honestly check on real battles whether it got better. The trainers run in the browser, free, no install; their models are built on our measurements, and everything simplified is marked in the game itself (“Honest about the model” button).

The training loop:

  1. Battle analysis finds a mistake that repeats from battle to battle and costs the most damage (“What to train” tab).
  2. The coach links to the right scene of the right trainer and gives a one-line rule for battle.
  3. The player sets a “goal for 10 battles” and plays as usual.
  4. On the next real battles the coach compares the share of bad battles with before; the “helped / didn’t” verdict comes after 30 battles. If it did not help, we revise the “mistake → exercise” link rather than praise the player.

Rules to keep it honest: advice only from many battles (one battle proves nothing); no more than two pieces of advice at a time; the main check is real battles, not trainer scores.

Where the trainer numbers come from. Server reticle delay and the cost of a flick come from our own measurements over thousands of shots. The distribution of shells inside the dispersion circle comes from WotStat statistics over 125 million shots — the most accurate source of shooting data in the game that we know of (half of the shells land within 0.37 of the radius from the centre, fewer and fewer towards the edge).

Celivalnya

Fig. 19. Celivalnya: scene, network and gun settings; on the right, the shot breakdown.

Trainer Status What it teaches Which mistake from the analysis it treats
Celivalnya live (prototype 0.1) shooting with real mechanics: the shell leaves from the server reticle. Scenes: single target, target switching, moving target, edge (armour angles), blind; “Training” and “As in game” modes; ping, network, gun and distance are adjustable flicks, unaimed shots, too little lead, non-pens, early deaths when peeking
“Angle from the side” duel live an angling duel: E 100, Maus, T110E3 and other heavies side exposed to the enemy, bad hull turns
Tankovalnya in development watch the enemy barrel and reload, show armour rather than side; peek and pull back on reload “side to an enemy at your nose”, “hull turn in front of a loaded enemy”, “exposed lower plate”, “nose up after pulling back”
Vyzhivalnya in development playing LT: route outside enemy view range, shooting from behind bushes and from beyond 445 m, shot and pull back early deaths under several barrels, “the shot gave you away before the lamp”, “shot from a bush”, “spotted while finishing off”
Artillery mode planned, name not chosen top view, full aim, leading moving targets given the long flight, target choice artillery misses on moving targets (the ellipse and target motion are already in the report)
Kustovalnya postponed camouflage of long-range TDs behind bushes spotted after firing from a bush closer than 15 m, spotted on the move

Where we are taking “analysis → trainer”

What TankCoach does not do, and what’s next

What there is not and will not be:

Test period. Everything is free; features marked “premium” are open to everyone for now. The server keeps the last 20 reports, older ones can be downloaded; battle statistics are kept.

In progress and planned:

Project news: Telegram t.me/tankcoach. The project changes almost every week; this description is as of 14.11.2026.

A young project: how to report a bug or suggest an idea

TankCoach is a very young project: the first versions appeared in autumn 2026, and it changes almost every week. It has bugs: a shot verdict may be wrong, a card inaccurate, a report may fail to build. Your messages are what helps us find and fix them.

Where to write:

If the mod or a report crashes, a technical error report is sent to the developer automatically (this can be turned off in the mod settings). We read every message; the best ideas from players have already made it into the report and the plans.