How strong is a counter pick? The numbers
Let’s take two heroes people most often call “annoying” and look at their worst matchups on Dotabuff [1] [2]:
| Matchup | Win rate in matchup | Overall win rate | Matches |
|---|---|---|---|
| Meepo vs Earthshaker | 44.88% | 52.95% | 28,458 |
| Huskar vs Ancient Apparition | 38.70% | 45.76% | 28,846 |
| Huskar vs Viper | 41.16% | 45.76% | 48,510 |
| Huskar vs Wraith King | 38.68% | 45.76% | — |
Meepo’s Dotabuff “Disadvantage” is 8.38% vs Earthshaker, 8.40% vs Elder Titan, 8.04% vs Winter Wyvern and 8.00% vs Underlord. But for most of his matchups it is under 3% [1]. Even in his best matchup (vs Anti-Mage), Huskar wins only 51.06% [2].
That gives a rule: most matchups barely matter, and a few specific ones matter a lot. A good last pick is one that lands one of those hard matchups on the most dangerous enemy.
Why the last pick is so valuable
The last pick has one big advantage: information. You see all five enemies and can counter the one that matters most to them. The first pick has to take a strong hero blind.
In pro play it is a bit different. At TI 2024, the team with first pick won 58.8% of 51 playoff games, and 63.6% of 22 games at the Royal Arena [3]. There, grabbing the strongest hero of the patch matters more. But even pro drafting is an information game: in game 2 of the grand final, Liquid left Naga Siren open on purpose (she was banned in the first phase in 48 of 51 playoff games) and made GG take her [3]. In pubs, the last pick simply has more information than the first, so use it.
Synergy or counters: which matters more?
Both. According to a 2017 review of studies, a model built on the sum of individual win rates got 63% accuracy, and adding hero pairs raised it to 72.9% [4]. Those pairs include both allies (synergy) and enemies (counters). The HSE study of 5 million matches also found that models which account for hero interactions predict better than simple ones [5].
The flip side: if you only look at a hero’s average win rate, the prediction is almost useless. In a 3.58-million-match benchmark, averaging win rates predicted results worse than “always bet on Radiant” [7]. So check your last pick both ways: who it plays against, and who it plays with.
A “strong hero” depends on your rank
Another reason not to trust tier lists: the same heroes perform differently at different ranks. OpenDota data from 19 Oct 2026 [6]:
| Hero | Herald | Divine | Change |
|---|---|---|---|
| Bristleback | 53.8% | 45.4% | −8.4 |
| Ogre Magi | 53.6% | 48.5% | −5.1 |
| Shadow Fiend | 45.7% | 40.7% | −5.0 |
| Bane | 43.3% | 50.3% | +7.0 |
| Visage | 48.6% | 55.6% | +7.0 |
| Enigma | 48.2% | 56.1% | +7.9 |
| Shadow Demon | 37.3% | 47.0% | +9.7 |
Bristleback is one of the best heroes in Herald and one of the worst in Divine. Shadow Demon is the opposite. The gap between ranks here is bigger than the effect of most counter picks. So “pick what the pros pick” can work against you at your rating.
A step-by-step last pick
- Find the main threat. Which enemy hero decides the game? In Outdraft it is marked with ⭐. Does your team already have an answer? If not, that is priority one.
- Check the lanes. Which of your lanes is Hard or Very hard? The last pick should help it, or at least not lose its own lane.
- Don’t take the weak link. A hero that ends up 🔻 weakest in this draft is a bad choice, even if it is meta.
- Pick from your pool. Hero consistency is linked to better results. A counter you play confidently beats a “perfect” hero you have three games on.
- Check the total. See how the overall draft chance changed. If the pick barely moves it, there may be a better option.
Common last pick mistakes
- Countering a hero that is already weak. If an enemy hero is the 🔻 weakest in this draft, countering it wastes your pick. Look at the ⭐.
- Picking from a tier list. A meta hero with no link to the draft gives less than a hard matchup on the main threat. And at your rank it may not even be strong.
- Forgetting your own lane. The last pick has a lane too. If it loses that lane Very hard, the counter pick bonus can be gone before minute 10.
- Picking a hero you don’t play. A couple of percent from a table is not worth learning a hero in a ranked game.
How it looks in Outdraft
During hero selection, Outdraft shows each hero’s impact in % and a lane rating: Good Even Hard Very hard. ⭐ marks the most influential hero, 🔻 the weakest, and at the top you see the overall chance, like “Dire 64%”. Pick suggestions take your role, the enemies and your allies into account, and are based on matches from the last four weeks, so they fit the current patch.
If the enemy ⭐ is on Huskar and your team has no healing reduction and no break, the best last pick is almost always the hero that fixes that. Even if it is not the strongest hero of the patch.