Top Players In History According To Davo A Deep Dive

Why Davo's Historical Rankings Actually Matter

Most people treating football analytics like a hobby stumble over David Odhiambo's work without realizing what's actually happening under the hood. The core problem isn't the data itself — it's understanding what metrics are being optimized and why certain players keep appearing at the top regardless of era. When I started digging into his methodology last year, I spent about three weeks just mapping out how he weights different performance categories across decades. The initial frustration came from realizing his system doesn't normalize for pace of play the way traditional stats do, which means raw numbers from the 1970s look inflated compared to modern equivalents unless you apply a correction factor himself built into the model. The first thing you need to understand before downloading any dataset or trying to replicate his rankings is that Davo uses a proprietary point system combining expected goals, progressive actions, defensive contributions, and consistency over time. Unlike FIFA's official historical rankings which rely heavily on awards and trophies, his model strips those out and focuses purely on measurable on-pitch output normalized per 90 minutes. I ran into a specific edge case while cross-referencing his top ten for the 1990s against Opta's archived data — the discrepancy showed up around midfielders because Opta didn't track progressive passes consistently before 2001. Davo's workaround was estimating those values using team possession averages and pass completion rates from match reports. That estimation error margin sits at roughly plus or minus 8 percent for pre-2001 midfielders, which matters if you're trying to rank players within two points of each other. The available datasets break down into three tiers. The free tier gives you the published rankings and basic player profiles. The intermediate tier, which costs around forty dollars annually, includes the raw season-by-season metrics with era adjustments already applied. The full research dataset runs about two hundred dollars and comes with the weighting formulas and documentation needed to build your own variant. I recommend starting with the intermediate tier even if you plan to go deeper later, because the era adjustment alone saves you weeks of manual recalibration work. A friend of mine who went straight to the full dataset spent six weeks before realizing the baseline assumptions were different from what he expected — he had to rebuild his entire comparison table because his control group wasn't accounting for the defensive handball rule change in 1992.

Here's the counterintuitive part that most people miss: Davo's model heavily favors players with long career spans at peak level over players with shorter peaks, even when the shorter peak involves objectively higher per-game numbers. Cristiano Ronaldo and Lionel Messi rank extremely high not just because of their output but because their consistency across fifteen plus years compresses the variance in the model. Players like Eric Cantona or Alfredo Di Stefano, despite incredible peak dominance, fall further down the list because their active years were concentrated and smaller in sample size. This creates an interesting tension where the model implicitly values durability over explosiveness, which some analysts argue disadvantages attacking players from eras with less rigorous scheduling. The practical workflow for anyone doing serious historical comparison starts with pulling the dataset, then building a controlled comparison set. I usually select five players from each decade spanning 1960 to present, making sure to include at least one goalkeeper and one defender per era so the positional adjustments don't skew the results. The adjustment factors Davo provides run about fifty different parameters depending on position, league strength, and era. Applying them correctly takes time — a typical full recalibration of the top hundred historical players across all positions runs somewhere between four to six hours on a decent machine if you're doing it manually, though the scripts he distributes can cut that to under an hour once they're running properly. One common mistake I see repeatedly is people comparing raw points across eras without applying the time-decay multiplier. The model assumes a depreciation rate of roughly 3 percent per calendar year for historical relevance, which means a player ranked highly in 1974 gets their point total reduced when viewed in 2024. This isn't about diminishing quality — it's about recency bias in the data availability and reporting accuracy. Without that adjustment, you end up crediting older eras with more reliable statistics than they actually had, which systematically inflates players from before 1985. The fix is straightforward once you know it exists: always run the time-weighted variant if your analysis includes pre-1990 players.

The biggest limitation of Davo's approach is that it cannot account for qualitative factors like leadership, big-game performance, or tactical adaptability. You will find players who were genuinely transformative in ways the model completely misses because those qualities don't translate into tracked metrics. I encountered this clearly when comparing Marco van Basten's actual historical impact against his model ranking — he falls outside the top thirty largely due to injury-shortened career length, which feels like a poor representation of what he was when healthy. For that reason I always run a parallel manual review alongside the quantitative output, looking at match footage and contemporary analysis to catch what the numbers smooth over. It adds maybe three hours of work per player you're evaluating but prevents you from drawing obviously wrong conclusions. If you're looking to get started, the download is available directly through his official analytics page at davofutbol.com/data. The documentation there is sparse but functional, and the community Discord has people who've already solved most of the common setup problems. Be prepared for the initial configuration to take a few hours if you're on a non-standard system, and don't skip reading the notes on data gaps for South American leagues before 1995 — the coverage is thin and the model fills those gaps using regional performance averages, which introduces its own layer of uncertainty.