Deltawall Academy

One tile per graduated token: ticker, venue and age, a 0–100 score, price change, then volume, liquidity and market cap for the chosen window.

buyBalance = 100 − |buyShare% − 55| × 2.5        weight 35
depth      = (liquidity ÷ MC) ÷ 0.15 × 100       weight 35
tradeSize  = (volume ÷ trades) ÷ $50 × 100       weight 30
score = Σ(part × weight) ÷ Σ(weight of measured parts)
final = min(score, activity cap)

Inputs

Live DEX pair: buys, sells, volume, liquidity and market cap for the selected window.

Limits & edge cases

Each part is clamped 0–100. Missing parts drop out of the divisor instead of counting as zero. Activity caps: volume < $10k/h → max 40; < $25k/h → max 55; < 60 trades/h → max 50; bots > 50% of trades → max 45; liquidity < 10% of MC → max 60; volume fallen below 30% of its 24h hourly average → max 50. The lowest cap wins.

How to read it

This is why a token with $2k volume can no longer show 98. A 55% buy share scores highest: balanced but leaning to buyers.

Chain and launchpad filters, timeframe, Graduated vs Trending, the live stat bar and the highest/lowest scores right now.

Trending: priceChange(window) > 0  AND  volume(window) > 0
          sort by trades desc, then volume desc
Hidden:   change(6h) ≤ −80%  OR  change(24h) ≤ −80%  OR  volume = 0

Inputs

Same live pair data as the tiles; up to 90 tokens.

Limits & edge cases

Only Solana is live; other chains show SOON and never show placeholder tokens.

How to read it

Fame / Shame rank tokens by score — never individual traders.

The token's overall score: the average of the four engine scores, then capped by activity.

integrity = min( mean(E1, E2, E3, E4 measured), activity cap )

Inputs

The four engine scores below plus pair volume, trades and liquidity.

Limits & edge cases

An engine with no measured data is left out of the average. The red note lists every cap that applied.

How to read it

75+ high integrity · 50–74 moderate · 25–49 caution · under 25 high risk.

Share of dollar volume from wallets that look like real, independent traders.

organic% = V(organic) ÷ (V(organic)+V(insider)+V(unknown)+V(bot)) × 100
shares rounded so the four add up to exactly 100.0

Inputs

Recent swaps with wallet, side, size and block.

Limits & edge cases

Unknown volume is never counted as organic.

How to read it

Under 50% means most of the volume shown on DEX screeners isn't real demand.

How much sellable liquidity stands behind the market cap.

backing% = liquidity ÷ market cap × 100
ring = min(100, backing% ÷ 20 × 100)   → 20% backing fills the ring

Inputs

Pool liquidity and market cap.

Limits & edge cases

Under 10% also caps the integrity score at 60.

How to read it

$1.6M MC on $40k liquidity (2.5%) means a few sells can move price hard.

Running total of buy dollars minus sell dollars through the window.

CVD(t) = Σ (+buyUsd or −sellUsd) for every trade up to t

Inputs

Each swap's side and dollar size, in time order.

Limits & edge cases

Follows the 5m / 15m / 1h / 24h buttons. No trades → empty chart, not a flat fake line.

How to read it

Price up while the line falls = the move isn't backed by net buying.

The measured figures behind the score (Flow / Risk / Execution) and a five-axis radar.

Organic    = organic%
Depth      = min(100, backing% ÷ 20 × 100)
Dispersion = 100 − top-10 holders %
Execution  = 100 − failed-swap rate% × 2
Retention  = buyers still holding %

Inputs

Same values as the engines.

Limits & edge cases

An unmeasured axis shows — and collapses to the centre.

How to read it

A lopsided shape shows which pillar is dragging the score down.

Wallets grouped by how many trades they made: 101+, 51–100, 21–50, 3–20, 2 and 1.

tier share% = Σ volume of wallets in tier ÷ total volume × 100
each wallet is in exactly one tier

Inputs

Trades grouped by wallet.

Limits & edge cases

Wallet count sits under each column; volume share on top.

How to read it

A few 100+ wallets holding most volume = machines. Many 1–2 trade wallets = broad participation.

Engine 01 score built from organic share, unique buyer ratio and dust.

E1 = 0.40·organic% + 0.35·min(100, AUBR ÷ 0.70 × 100) + 0.25·(100 − dust trades%)
if bot trades > 50% → E1 = min(E1, 45)

Inputs

Classified swaps for the window.

Limits & edge cases

Parts that can't be measured drop out of the weights.

How to read it

A cap note means the score was held down, whatever the other parts say.

Every trade lands in exactly one bucket, shown by dollars and by trade count.

Bot      : trade < $10  OR wallet has ≥2 buys AND ≥2 sells (wash)
           OR wallet bought and sold in the same block (sandwich)
Insider  : wallet funded from a shared SOL source (cluster)
Unknown  : wallet with 4+ trades that isn't wash or sandwich
Organic  : everything else
ratio    = organic ÷ bot   (dollars and trades separately)

Inputs

Wallet, side, size and block of each swap; funding clusters from Engine 04.

Limits & edge cases

Rules run top to bottom; bot wins over insider. No bot side → shows “no bot”, not ∞.

How to read it

Bots are often small in dollars but huge in count — always compare both charts.

How many organic wallets actually bought.

AUBR = unique organic buyers ÷ unique organic wallets
score = min(100, AUBR ÷ 0.70 × 100)

Inputs

Organic trades only.

Limits & edge cases

No organic wallets → —.

How to read it

0.70+ means most real wallets are buying. Low means the same wallets are trading back and forth.

Count of trades in size buckets from under $10 to over $1k.

dust trades% = trades < $10 ÷ all trades × 100

Inputs

Swap dollar sizes.

Limits & edge cases

Trades under $10 are always treated as bot.

How to read it

A tall first column = spam. Healthy demand spreads into the $50–$1k range.

Engine 02 score from absorption, buy pressure and churn.

E2 = 0.40·absorption + 0.35·buy pressure + 0.25·(100 − churn%)

Inputs

Organic trades grouped by wallet.

Limits & edge cases

Unmeasured parts drop out.

How to read it

High = organic money is arriving and staying.

Organic buy dollars that are still held vs already re-sold.

absorption = held ÷ (held + re-sold) × 100
only pending (too new) → 50

Inputs

Each organic buy, checked against later sells by the same wallet.

Limits & edge cases

Pending buys are excluded until there's enough time to judge.

How to read it

Low absorption = buyers flip quickly; the volume doesn't stick.

Share of organic dollars on the buy side.

buy pressure = organic buy $ ÷ (organic buy $ + organic sell $) × 100

Inputs

Organic trades only.

Limits & edge cases

No organic trades → —.

How to read it

Under 50% while price rises is a warning.

Cumulative organic buy-minus-sell, plus wallets trading both sides.

churn% = wallets with buys AND sells ÷ all wallets × 100

Inputs

Organic trades in time order.

Limits & edge cases

Follows the window buttons.

How to read it

High churn = the same money circling, not new money.

Engine 03 score from failed swaps, exit cost and who is failing.

E3 = 0.40·(100 − rejection%) + 0.40·exit score + 0.20·organic share

Inputs

On-chain failed swaps (error 0x1771, slippage exceeded) and off-chain Jupiter route simulations for exit depth. Quotes never fail; only executed transactions do.

Limits & edge cases

Unmeasured parts drop out.

How to read it

Low = people can't get trades through or exits are expensive.

Share of swap attempts that failed.

rejection% = failed ÷ (failed + successful trades) × 100

Inputs

Failed transactions on the pool, split slippage / gas / other.

Limits & edge cases

Nothing indexed → —, never 0%.

How to read it

Many failed sells = people struggling to exit.

Failures from real wallets vs bot retries.

organic share = organic failures ÷ (organic + bot-retry failures) × 100

Inputs

Failed swaps with their wallets.

Limits & edge cases

Bot retries are counted separately.

How to read it

Organic = signers not flagged as rapid retries (≥3 failures/min); we cannot verify humans. Organic failures hint at real demand or panic; bot retries are noise.

Live Jupiter quote: how much price moves if you sell $1k, $5k, $10k.

exit score = 100 − average impact% × 15   (clamped 0–100)

Inputs

Live Jupiter quotes.

Limits & edge cases

Check limit: $10k sell ≤ 15% impact.

How to read it

About 6.7% average impact already scores zero.

Priority fees and tips — who pays to get in first.

bot fee share = bot fees ÷ all fees          avoid > 70%
recipient overlap = shared recipients ÷ all    avoid > 25%
burned = fees on failed txns ÷ all fees        avoid > 20%
caution starts at 75% of each limit

Inputs

Parsed Solana transactions.

Limits & edge cases

Can't compute → “not computed”, never 0%. Limits are guides, not backtested.

How to read it

Rising fees can mean bot competition, not demand.

Engine 04 score from retention, buyer concentration and holder spread.

E4 = 0.30·retention + 0.30·top-5 score + 0.40·spread
spread = mean( min(100, Nakamoto × 10), 100 − cluster supply% × 2 )

Inputs

Holders, funding clusters, window buyers.

Limits & edge cases

Unmeasured parts drop out.

How to read it

Nakamoto = fewest wallets that together hold over half the supply.

Window buyers still holding.

retention% = buyers who never sold ÷ all buyers × 100

Inputs

Trades grouped by wallet.

Limits & edge cases

Check limit: early buyers ≥ 30% still holding.

How to read it

Low retention = buyers are using the token to flip.

Share of buy dollars from the five biggest buyers.

top5% = Σ top-5 buy $ ÷ all buy $ × 100
score = 100 − (top5% − 30) ÷ 50 × 100   (30% → 100, 80% → 0)

Inputs

Buy trades by wallet.

Limits & edge cases

No buys → —.

How to read it

Over 50% means a handful of wallets are the market.

Real vs dust holders by supply, transfer-only wallets and exchange deposits.

cluster supply% = Σ supply of wallets sharing a funding source
flag when any cluster > 10%

Inputs

Holder balances, SOL funding sources, transfers.

Limits & edge cases

Transfers show tokens and % of supply — never added to trading volume.

How to read it

A big holder sending to an exchange is a stronger exit signal than a small DEX sell.

Left: share of measured checks passed. Right: how many checks had data.

score% = passed ÷ (passed + failed) × 100
coverage = measured ÷ 26

Inputs

All 26 checks.

Limits & edge cases

Unmeasured checks never count as passed.

How to read it

89% on 9 of 26 checks is far weaker evidence than 89% on 24.

Six tiles: liquidity/exit, net flow, real buyers, bot share, concentration, move health.

bot share = max(bot $ share, bot trade share)
GO needs ≥ 3 real buyers, bot share ≤ 50%, non-negative flow
verdict = AVOID if any tile is AVOID

Inputs

Same measured values as the engines.

Limits & edge cases

No data → no data, never GO.

How to read it

One AVOID tile outweighs a high score.

26 checks in Liquidity, Demand, Safety and Ownership. Each row: status, name, value, limit, ⓘ reason.

Liquidity ≥ $10,000
Exit cost of $10k ≤ 15%
Early buyer retention ≥ 30%
Bot share ≤ 50%

Inputs

Live pool, quotes, trades and holders.

Limits & edge cases

Limits are guides, not backtested, not advice.

How to read it

Tap ⓘ on any row for why that limit exists.