Continue with Steam Use statistics to make a shortlist, not to declare a winner before you play. Start with the same rank and patch where the site offers those filters, check how many matches support a hero’s rate, and ask whether an item solves the problem you face. A high win rate is an observation about a group of games; it does not guarantee a result for you or prove that a purchase caused those wins.
The site’s hero and item lists use imported data from the community-maintained Deadlock API. Coverage and updates can vary. The examples below are hypothetical, not current site rankings or actual match results.
Compare heroes in the same context
Open the Heroes list. Its toolbar offers a rank picker and Patch, plus Role, Tier, search and Sort. The table shows win rate, pick rate and matches. Select a rank and patch relevant to your question, then compare rows under the same settings. The rank filter groups games by the match’s average badge, so it is an approximate match-level context, not a guarantee that every participant has your exact rank.
The Role control filters the site’s hero type labels; it is useful for narrowing the list, but does not prove a hero has only one possible job. The Hero Tier List offers the same broad rank, patch, role and tier controls. Its cards show win and pick rates but not the match count: return to the Heroes table under the same settings to check that count. Tier letters on these lists are derived from win-rate thresholds, not a separate measurement of impact.
A hypothetical hero comparison
For example, imagine Hero A has a 52% win rate in 200 matches and Hero B has 51% in 1,000 matches under the same filters. A’s higher rate is worth investigating, but the smaller sample can fluctuate more, and a one-point gap alone does not establish that A is better for you. Changing rank or patch changes the population, so a rate from another selection is not a like-for-like comparison. Check both heroes’ 
abilities and playstyle, then try the one that fits the job you want to perform. Pick rate shows how often a hero appears in the dataset; it is not an independent quality score.
Check sample and data limitations
The default is All ranks and the first available patch in the filter. Rank and patch options depend on imported snapshots. A fresh patch or narrow rank can have fewer observations; a larger sample generally reduces random variation, but it does not fix a biased or incomplete dataset. The site does not claim that its sample includes every match.
Missing rows and freshness
An Updated label, when shown, reflects an import or source-file timestamp. It is a useful freshness clue, not proof that every selected patch/rank snapshot is complete or refreshed at that moment. Check the selected patch after an update and compare like with like; don’t merge a prior patch’s rate into a new one as though the game had not changed.
Missing data also needs care. A hero without a row for a selected rank can appear with zero values; the display can represent a missing imported row, not a verified 0% win rate or proof that no one played the hero. An item can be absent because it is outside the list’s coverage or has no usable match count in the selected snapshot. If an entire item snapshot is unavailable, the page can fall back to imported item metadata, which may not match the selected rank and patch. A displayed zero or absent card is not by itself a measured zero result. Treat uncertain scope as a reason to withhold a precise comparison.
Use item data without assuming causation
Open the Item Tier List and select the available rank and Patch. Category narrows items to Weapon, Vitality or Spirit; Tier and search narrow the displayed cards. The Standard list covers eligible shop items and shows win rate and pick rate, but the cards do not show the match count. They are a starting point for investigation, not a complete evidence table.
An item’s win rate is calculated from wins and matches in its item-stat row; the snapshot’s pick rate uses players associated with the item relative to total players in that dataset. That is a different population from a hero’s match row. Do not compare an item percentage directly with a hero percentage as though they were measured over the same games.
Click an item to read its description and, where available, Current Statistics, including Matches and Players, and Recent Patch Changes. The item detail page uses imported item data and does not inherit the tier-list rank and patch selection. Its match count is therefore not necessarily the denominator behind the filtered card. Compare only when you can establish matching scope; otherwise note the limitation rather than pairing the two numbers.
Why item win rate is not item impact
Imagine Item X appears with a 56% win rate and Item Y with 50% in a hypothetical item dataset. X may be purchased more often by players who are already ahead or reach a later stage of a match; Y may be bought in difficult defensive situations. Heroes, purchase timing, opponents and player choices can differ too. For example, if X is usually bought late after a team has already gained an advantage, those favorable games select into X’s sample. These are possible explanations for selection bias, not facts established about X or Y.
The percentages describe associations among recorded purchases and outcomes. They do not isolate the effect of buying X instead of Y in otherwise identical games. Ask what each item does, whether you can buy and use it at the right time, and whether it addresses the threat in your match. A high rate is a reason to investigate, not a promise that changing one item will produce the same rate.
Should I always buy the item with the highest win rate?
No. A high rate can reflect who buys the item and when, including players already in a strong position. Check its function and whether it addresses your immediate problem; compare matching data scopes where available, then test whether you can use it effectively.
Compare the tools without mixing their scope
| Tool | Use it for | Scope warning |
|---|---|---|
| Heroes | Rank/patch rates and match counts | A selected-rank missing row can display zeros. |
| Hero Tier List | Scan hero tiers and rates | Cards omit match counts; tier is based on win rate. |
| Item Tier List | Rank/patch item rates and categories | Cards omit matches; missing snapshots can fall back to imported data. |
| Individual hero and item pages | Abilities, descriptions and available  stats |
The page need not inherit list filters. |
| Compare Heroes | Compare two to four heroes side by side | Aggregate/fallback data are not the selected list slice. |
Scope check: Are the patch, rank context, mode, data source and denominator known to match? If a card does not show a count or a profile does not expose that same slice, leave the comparison unresolved instead of attaching an unrelated count.
The comparison page is useful for exploring practical differences, but its hero statistics use an aggregate source with fallbacks. A number there and a filtered table number may describe different samples. Hero profiles may show a Meta Verdict with a rate and count, plus sections such as abilities or statistical builds when data exists; those sections are not automatically scoped to the filters you previously chose.
Make a decision, then recheck
- Define the task: for example, choose a hero you can use to protect teammates, or an item that addresses a recurring threat.
- On the relevant list, select the same available patch and rank for each candidate. Record the displayed rates and, for heroes, match counts. Note any unavailable or uncertain data rather than filling gaps with zero.
- Read the hero’s abilities or item description. For items, use the detail page’s count as context only, unless you can confirm it represents the same slice as the card.
- Try the choice in matches and observe whether you can execute its intended job. Use the replay-review guide to inspect one decision rather than judging from a single win or loss.
- Recheck after a patch or data refresh. If the sample or context changes, reconsider the shortlist; a tier letter is not a permanent recommendation.