The logic sounds reasonable: pos 3 creates a second front, forces enemies to respond in two places simultaneously, and converts individual skill into map control. In theory, split-push pos 3 should be viable.
The data says otherwise.
The Hard Floor
Teamfight participation below 0.3: 0.0% win rate in 13 games. Never won. Once. In 13 games.
Below 0.4: 5.6% win rate. Below 0.5: 25.6%. The cliff between 0.5 and 0.6+ is dramatic. At 0.6+ teamfight participation, the win rate jumps to 65%. This is the clearest single threshold in the entire dataset.
Tower Damage Comes From Winning Fights
Tower damage above 1k: 78.5%. Above 5k: 92.7%. Above 8k: 100%. These numbers look like tower damage wins games — but the causation runs the other way.
Teams that win teamfights then push towers immediately. Tower damage is an output of winning fights, not an input. An offlaner who pushes towers while their team fights 4v5 accumulates tower damage and loses.
The Lycan Paradox
| Hero | Win Rate | Avg Tower Damage | Note |
|---|---|---|---|
| Lycan (pos 3) | 20% | 6,047 | Pushes, team loses 4v5 |
| Terrorblade | 28.6% | 11,340 (highest) | Farm dependency |
| Nature's Prophet | 25% | High | Global split-push |
Lycan as pos 3: 20% win rate despite 6,047 average tower damage. The pattern is consistent across all split-push specialists — high tower damage, low win rate. The hero pushes towers while the team loses fights 4v5.
The Assists + Tower Damage Combo
Assists HIGH + tower damage HIGH: 86.2% win rate. Both LOW: 2.2%. The combination you need is both — fight participation and objective conversion. Not one or the other.
What Damage Taken Tells You (Nothing)
Average damage taken in wins: 40,446. In losses: 40,361. Statistically identical. Tanking damage predicts nothing about whether you win. Stop evaluating offlaners by how much they soak.
The Best Pos 3 in This Meta
Batrider: 66.7% win rate. Axe, Tidehunter, Centaur — all strong. The hero theme is clear: every top-performing pos 3 initiates teamfights. The split-push specialists are at the bottom.
The Data Behind This
All findings derived from DOTApulse’s 403-match pro dataset across patch 7.41.
Read the full research →Related articles