Do Solana Memecoin Strategies Work? 2,380 Tokens Tested

5½ months of data. 405,790 price snapshots. 2,371 final outcomes. Zero reliably profitable rule-based strategies.

The short version, for the people who will only read this far:

• We tracked 2,380 Solana memecoins from 19 December 2025 to 4 June 2026, taking 405,790 price snapshots along the way.
• 43.4% of every token that reached a $250k market cap ended as a rug. That is the single most common outcome. Not mooning. Rugging.
• We backtested more than 100 rule-based strategy permutations against the full dataset with realistic slippage. None produced a reliable positive expected value.
• The best result in the entire study was +0.3% per trade, on a sample so small it is statistically indistinguishable from zero.
• The honest takeaway: if your memecoin edge comes from public DEX data and anything slower than co-located infrastructure, the data says you are paying to play a game you cannot win.

Everything below is the work behind those numbers.

Why we built this

By late 2025 the Solana memecoin machine was running hot. Pump.fun was clearing roughly $2B in daily DEX volume at its Q1 2026 peak. Something like 11.6M new tokens were minted in 2025, of which under 1% ever "graduated" to a meaningful market cap. The top trading bots were reportedly pulling in seven figures a month in fees alone. And the timeline was wallpapered with screenshots of 100x trades.

So we asked a deliberately narrow question:

Is there a rule-based strategy, built only on publicly available DEX data, that produces consistent profit on Solana memecoins?

The word that matters there is public. We were not interested in the strategies that work for a small professional subculture: sub-50ms RPC latency, Jito MEV bundles, private Telegram alpha, copy-trading known winning wallets, sniping the first blocks before retail can react. Those edges are real. They are also unavailable to anyone trading from a normal machine with a normal data feed.

We wanted to test the thing most people actually have access to: ordinary execution, 30 to 60 seconds behind the public data, no insider feed. The version of memecoin trading that gets sold in YouTube tutorials and paid signal groups.

We had run versions of this before (a BSC honeypot scanner, a momentum grinder, a multi-profile bot) and every one showed the same pattern: the backtest looked promising, the paper trading came back negative. We suspected the problem was sample size. Weeks of data and a few dozen trades is not enough to tell signal from noise.

So this time we did something different. We built an observational dataset with no trading activity at all. Just watch what happens to tokens once they reach a given size, and track them across their lifespan. Once we had months of it, we could test any strategy we liked against real outcomes, with no look-ahead bias.

What we collected, and how

The collector is a single Python service running in a Docker container, 24/7 for five and a half months. It does two jobs.

It scans. Every minute it polls the DexScreener API for new Solana tokens crossing $250,000 in market cap. That threshold is a judgment call: it means a token has attracted enough attention to be interesting, but is not yet established. Below it, the noise is overwhelming. Above it, the move you care about has often already happened.

It tracks. For every token it discovers, it saves a snapshot every 10 to 15 minutes: price, market cap, liquidity, 1h and 24h volume, buy and sell transaction counts, holder data where available. Tracking continues for 48 hours, or until the token is classified as a rug.

Every token then gets a final outcome based on its price history:

We wait the full 48 hours before calling an outcome final, so we capture the entire first wave rather than a snapshot of it.

The result is a PostgreSQL database with two tables: one row per discovered token with its metadata and final outcome (2,380 Solana tokens, of which 2,371 reached a final outcome), and a time-series table of every snapshot (405,790 rows, roughly 170 per token). The whole thing is about 77 MB. Compact enough to throw at Pandas and answer almost any question in seconds.

The limitation we want to put up front, not bury

Here is the obvious objection, and we would rather raise it ourselves than have you raise it in the replies: by only tracking tokens once they pass $250k, we never observe the launch phase. The first seconds, where sniping bots and pre-graduation snipers extract their edge, are invisible to us.

That is true. And it is the point, not a flaw.

We are not claiming "memecoins are unwinnable for everyone." We are testing a specific, falsifiable claim: can you win with public data and realistic latency, the way it is sold to retail? If there is an edge in the first three blocks, it belongs to people with infrastructure you do not have. Measuring whether that edge exists would require a different dataset and would not tell you anything actionable, because you cannot act on it anyway.

So read everything below as an answer to the question retail actually faces, not the one a professional sniping desk faces.

What the data looks like

Outcomes across the 2,371 tokens that reached a final result

The headline finding lives in the first row. The most likely thing to happen to a Solana token that reaches a $250k market cap is that it rugs. Not that it moons, not that it drifts sideways. It rugs, 43% of the time.

The market does not change its mind

The moon rate sits between 24 and 27% month after month, with January's euphoria the only real outlier. This stability matters more than it looks. It means the market is not flipping between regimes. A strategy that fails in one month is not waiting for conditions to change. The conditions are the conditions.

Where the rugs live

Pumpswap, the successor to Pump.fun's bonding curve, is where four out of five tracked tokens live, and where nearly half of them rug. Raydium and Meteora rug far less often, but they also moon less often. There is no free lunch in the venue choice: lower rug rate comes bundled with lower upside.

Higher liquidity, lower upside

Tokens that already had more than $500k in liquidity when we found them mooned 1.8% of the time. They are too established to run again. The sweet spot for upside is $25k to $100k of liquidity, which is, predictably, also where the rugs cluster. The data keeps handing you the same trade: more upside, more risk, in lockstep.

Younger tokens moon more and rug more

A token under an hour old moons 31.6% of the time and rugs 62.9% of the time. The risk and the reward both peak at the same moment, which is exactly what makes the early window so hard to trade with a fixed rule.

How far the moons actually go

Of the 613 moons:

Almost half of all moons "only" doubled. The mega-moons everyone is actually chasing, a gain above 1,000%, accounted for 52 tokens out of 2,380. That is 2.2%. You cannot build a repeatable strategy around catching a 2.2% event, and the math of the next section is largely the math of that sentence.

What we tested, and what came back

We did not test one strategy. We tested the strategy space.

Discovery filters. The classic approach: find the combination of DEX, age, liquidity and market cap at discovery that predicts a moon. We tested the profiles that had looked promising on small samples (one had shown an "80% moon rate" on ten tokens). Run against the full 5½ months with a realistic 10% round-trip slippage, every filter returned between -5% and -15% per trade. The earlier "80% moon" results were noise on samples of ten to thirty.

Trajectory features. Maybe how a token moves in its first one to three hours predicts where it ends up. We entered at the second snapshot and measured forward returns across seven buckets of early movement. Under chronologically correct simulation, all seven were negative. The best, tokens that stayed flat early, still lost 10% per trade.

Conditional edges. Maybe single features fail but combinations work. We searched more than thirty interaction subsets: DEX by early-move by liquidity tier, volume acceleration by early movement, and so on. Zero produced positive expected value after slippage. The best find lost 2% on a sample of 33, which is to say it found nothing.

Time of day and day of week. Asia, EU and US sessions moon at 28.7%, 25.9% and 23.7% respectively. Thursdays run 23.4%, Saturdays 28.1%. These differences exist, but they are far too small to trade against slippage.

Exit timing. For tokens that reached a 50% gain, the median peak came 9.4 hours after entry. For tokens that reached 100%, 12.5 hours. Only 15 to 24% peaked within the first three hours. This is the quiet killer: quick-flip strategies that take profit within an hour or two miss most of the actual runs, but holding longer walks you straight into the dump, which usually arrives within 24 to 48 hours. The window between "too early" and "too late" is narrow and moves around.

Trailing stops. We brute-forced 100 combinations of activation, trail and hard stop across eight subsets. The single best result in the entire study:

Meteora-only, activate at +50%, trail 15%, hard stop 10%: +0.3% per trade under realistic tier-based slippage, on 195 tokens.

That is it. That is the peak of the mountain. +0.3% on a sample of 195, which collapses back to -1.5% the moment you model slippage as a flat 10% instead of a tiered estimate. It is, in every meaningful sense, zero.

Why it does not work

The null result is not bad luck. There are structural reasons, and they reinforce each other.

Selection bias in the public data. By the time a token is visible on DexScreener at $250k, it has already passed through algorithmic discovery, copy-trading bots and a first wave of trading. The real edges live in the seconds before that, which public feeds do not show you.

Slippage eats the margin. Realistic round-trip slippage on these tokens runs from 8 to 12% in the $50k to $100k liquidity band down to 2 to 3% above $500k. If a strategy shows a +5% average gain before costs, realistic execution turns that into a 3 to 7% loss. That gap is the entire ballgame, and it is the gap we measured again and again.

The outcome distribution is bimodal. Tokens are either rugs and dumps (down 50 to 95%) or moons (up 100 to 5,000%). The middle is thin. Tight take-profit and stop-loss rules miss the whole moon distribution, so you collect small wins and eat the big losses. Wide rules have a low win rate that slippage then destroys. You need a strategy that hits its stop often without ever getting killed, and catches moons most of the times they happen. That is mathematically hard to build, and we could not build it.

The timing is adversarial. Moons peak late (9 to 15 hours). Rugs collapse fast (1 to 3 hours). A trailing stop set tight enough to dodge the rug triggers on the first healthy pullback. Set loose enough to ride the moon, and it has already absorbed the rug damage by the time it fires.

You are not the smartest money in the pool. The overwhelming majority of bot transactions on Solana fail to slippage, MEV and sandwiching. The ones that succeed run on infrastructure (low-latency Geyser feeds, Jito bundles, co-located nodes) backed by real research budgets. From a home machine on public data, you are the liquidity, not the predator.

The one thread that is not entirely dead

We will not pretend the Meteora result is nothing, but we will not let it become a cliffhanger either.

The +0.3% on Meteora is statistically weak. But it is interesting because it points at a mechanism rather than a coincidence. Meteora uses a dynamic liquidity market maker rather than a standard AMM, it sees less sniping-bot competition than Pumpswap, and its LP structure can give better execution in certain windows.

If there is an edge there, it almost certainly does not live in rule-based entry filters. It lives in LP positioning (range orders that earn fees), in timing around migration events, and in spotting tokens that graduate from Pumpswap to Meteora. That is a completely different class of strategy from the one this study tested. It is worth investigating. It is not a result we can claim today.

What a null result is actually worth

We did not find a profitable strategy. We found something more durable.

As evidence. "5½ months, 2,380 tokens, 100+ strategy permutations, zero reliably positive expected value." That is honest empirical pushback against an industry of paid signal groups and YouTube traders selling the opposite. Negative results almost never get published, because nobody can monetize them. Which is exactly why they are worth publishing.

As a baseline. For future work with machine learning or richer on-chain features (holder concentration, dev wallet patterns, smart-money flow), this dataset is a calibration floor. If a model cannot beat -5% per trade on this data, it has not found a real edge. It has found a way to overfit.

As ground truth. A stable 25% moon rate every month. A constant 43% rug rate. A clean liquidity-versus-age trade-off. A bimodal peak distribution. These are quantified facts about how the Solana memecoin market actually behaves, as opposed to anecdotes about the one trade that went 100x.

What this means if you are still trading these

We will be blunt, because the data is.

1. Stop autotrading rule-based memecoin strategies on public data. The dataset says it does not work, and it says so consistently.
2. Use the system as a monitor, not a trader. Let it surface interesting candidates and leave the entry decision to a human who knows something the data does not.
3. If you have external alpha, automate the discipline, not the decision. A bot is good at executing a clean exit without flinching. It is not good at finding the trade.
4. If you keep experimenting, look at on-chain features and the Meteora LP angle, not at another permutation of entry filters. We have already searched that space for you. It is empty.

The uncomfortable summary is that the most profitable thing we did across five and a half months and 405,790 snapshots was to not trade, and to write down why.

Logs over lambos.

Frequently Asked Questions

Do Solana memecoin trading strategies actually work?
Based on 5½ months of data across 2,380 tokens, no rule-based strategy built on public DEX data produced reliable positive expected value once realistic slippage was modeled. The best result was +0.3% per trade on a sample of 195 tokens, which is statistically indistinguishable from zero.

What percentage of Solana memecoins rug?
43.4% of tokens that reached a $250k market cap ended as a rug. This is the single most common outcome. Another 11.6% dumped without a hard liquidity rug, 19.2% drifted sideways, and 25.9% reached at least a 100% gain.

How often do Solana memecoins moon?
Roughly 1 in 4 tokens that cross $250k market cap will reach at least a 2x gain at some point. The moon rate was stable between 24% and 27% month after month. Mega-moons (over 1,000% gain) accounted for only 2.2% of all tracked tokens.

Why don't memecoin trading bots work for retail?
Three structural reasons: realistic round-trip slippage (8-12% in low-liquidity tiers) eats the margin, the outcome distribution is bimodal so tight rules miss moons while loose rules absorb rug losses, and the real edges live in pre-graduation seconds that public DEX data does not show.

Is there any edge in trading Solana memecoins?
The study found one weak signal on Meteora-only tokens (+0.3% per trade) but called it statistically inconclusive. Any remaining edge likely lives in LP positioning, migration timing, or smart-money wallet patterns - not in rule-based entry filters on public price data.

Methodology notes: entries simulated at the second snapshot (1 to 3 hours after discovery) to avoid look-ahead bias. Exits resolved chronologically on the first snapshot meeting the stop or target. Baseline slippage modeled at a flat 10% round-trip, with a tiered alternative from 2% (over $500k liquidity) to 12% (under $50k). DexScreener API rate limits constrained sampling to one snapshot per 10 to 15 minutes rather than real time. The holder-count field returned inconsistently and was excluded as a feature. The $250k discovery threshold excludes the pre-graduation phase by design.