libMEV - ELO Based Searcher Ranking

Searching is a game that requires a combination of skill, strategy, persistence, and of course, luck. While many players and platforms currently ultilize total profits as a ranking metric, we believe that this approach is fundamentally flawed. In this blog post, we introduce a new ranking algorithm inspired by the ELO rating system used in chess and explain why it is a better alternative.

Problems With Total Profits

Total profits is a commonly used metric to rank searchers. On the surface, this seems like a straightforward and logical approach. After all, a searcher who consistently wins money should be ranked higher, right? However, there are several fundamental flaws with this approach:

• Short Term Variance: Searching involves navigating highly dynamic and often unpredictable markets. As a result, searchers can experience significant short-term variance in their earnings due to market conditions. For example, having an optimized arbitrage bot might yield substancial profits during periods of heightened speculation (a.k.a shitcoin season), while the same bot might be overshadowed by a liquidation-focused bot during periods of volatility.

• Timing and Luck: Being in the right place at the right time can significantly impact a searcher's profits, regardless of their skill level. A lucky early entry into an uncompetetive ecosystem can result in substantial profits that are not necessarily indicative of skill.

LibMEV Searcher ELO Ranking

Inspired by the ELO rating system used in chess, we propose a new searcher ranking algorithm that addresses these limitations and provides a more nuanced evaluation of a searcher's skill and performance. These are the properties of the modified searcher ELO rating algorithm:

• Starting Rating: Every searcher begins with a certain number of points.

• Winning and Losing: The amount of points exchanged depends on the difference in their ratings. If a lower-rated searcher beats a higher-rated searcher, they gain more points than if they beat someone with a similar rating. The searcher is considered to have "won" if they've accumulated more profits than the other searcher (in a certain time period).

• Expected Outcome: The system calculates the expected outcome of the game based on the searcher' ratings. If a searcher with a higher rating wins, it's not a big surprise, so they gain fewer points. But if the lower-rated player wins, it's a bigger upset, so they earn more points.

• Continuous Process: er each time period, the ratings are adjusted accordingly. This process repeats every interval, as long as the searcher has sent a bundle, continuously updating the ratings based on the outcomes.

Sneak Peek

The video below depicts the ELO rating of the top 20 searchers (by best ELO score) from the merge, till Nov 24th 2023.

searchers = [
    "0x6b75d8af000000e20b7a7ddf000ba900b4009a80",
    "0x000000d40b595b94918a28b27d1e2c66f43a51d3",
    "0x00000000003b3cc22af3ae1eac0440bcee416b40",
    "0x000000000035b5e5ad9019092c665357240f594e",
    "0x64545160d28fd0e309277c02d6d73b3923cc4bfa",
    "0xd4674001a9a66b31f3c09e3b1fec465404c83d35",
    "0xb0bababe78a9be0810fadf99dd2ed31ed12568be",
    "0x67fd18ffedb11f820e54b43d54fa756d74f6add7",
    "0x0000cd00001700b10049dfc947103e00e1c62683",
    "0x00fc00900000002c00be4ef8f49c000211000c43",
    "0xcb3702bc25b0f284b032e5edf1a1ebea2fe43255",
    "0x493f461aead031cee2027f1b95370a692611acb9",
    "0xb0000000aa4f00af1200c8b2befb6300853f0069",
    "0x53facee52e897740b140f5304e9cd9dc6238d735",
    "0x9db7378614d8d9d7149c4ee4763f88c38f9b1517",
    "0x95c0c8a41caaa33f22ff042dc4279ed2b4e74041",
    "0xe545c3cd397be0243475af52bcff8c64e9ead5d7",
    "0x00000000de337b4fff5fcbe4df67a85d0bad5d16",
    "0x000000adf684e05b0e33f8724fff2e9ce8cc6248",
    "0x5db209f7f4cbd7011b486b8d99e5ae616947a138",
    "0x24902aa0cf0000a08c0ea0b003b0c0bf600000e0",
    "0x00000000009e50a7ddb7a7b0e2ee6604fd120e49",
    "0xe8c060f8052e07423f71d445277c61ac5138a2e5",
    "0x00004ec2008200e43b243a000590d4cd46360000",
    "0x0000e0ca771e21bd00057f54a68c30d400000000",
]

From the video above, we can deduce that 0xb0babae was probably the best searcher in around early Oct 2022, as they were ahead of the competition by a good margin (probably due to them being an early searcher/builder). However they couldn't hold onto that lead and eventually fell down the rankings.

The next few months, we saw relatively fierce competition amongst the top searchers, there wasn't a clear champion, until mid Jan 2023, where 0x0000...416b40 started to breakaway from the rest of the competition, and stood there for ~3 months.

0x0000...cc6248 overtook the crown and briefly became the champion for a few days, until...

0x6b75d...09a80 (jaredfromsubway.eth) came into the scene and has been the #1 searcher since.

Why?

Our hope is that this new ELO ranking system can bring more insight into the searchers of the ever changing dark forest, better reflecting the skill of searchers across generations, and market conditions.

When?

LibMEV searcher ELO ranking will be released in libMEV shortly. Stay tuned!