totoonsol

Data Science @ ETH Zürich
Solana dev - on-chain data & memecoin market mechanics

[totoonsol @totoonsol_]

paper trading teaches you a market that doesn't exist.

small PSA for anyone practising memecoins on a
simulator before putting real money in. the fills
you're seeing are not what you would have gotten on
chain, and i measured exactly how far off they are.

pumpfun's bonding curve is a constant product pool.
it starts at 30 virtual SOL and 1.073b virtual tokens.
two things happen on every buy that most paper models
skip entirely:

1. a 1.25% fee (0.95% protocol + 0.30% creator) is
taken before your SOL ever reaches the curve.

2. your own order moves the price while it executes.
you get T * x / (S + x) tokens, not x / (S/T). the
bigger your size relative to the pool, the worse your
average fill.

i implemented this from the program IDL and validated
it against 119 real trades across 3 tokens at different
points on the curve. tokens, net SOL and fees match to
the lamport on all 119. so the numbers below aren't a
model of a model, they're the chain.

worked example, real buy at slot 448,107,312:

1.782 SOL in
actual tx received 30,636,219.502858 tokens
my implementation 30,636,219.502858 tokens
displayed price minus nothing 32,314,949 tokens

that's 5.48% more tokens than the chain would give you.
1.25% is the fee, the rest is your own impact.

and it's not a constant offset, which is the part that
matters. on a fresh curve a 10 SOL buy overstates your
fill by about 26%. on a curve near migration the same
buy is off by about 10%. a 0.1 SOL buy is off by 1.5%.
so the error scales with exactly the behaviour new
traders are trying to learn: size up, enter early.

selling has the same problem in reverse. paper tools
show unrealised pnl at the displayed price, but exiting
a large position pushes the price down as you sell. the
number on your screen is not what hits your wallet.

why this matters: a lot of people practise on +5 to +15%
moves. a strategy that looks profitable on paper can be
flat or negative live purely from execution, before you
even count priority fees, jito tips, failed transactions
and account rent.

i'm not selling anything and i'm not naming any tool.
the ones out there are genuinely useful for practising
workflow and discipline. they just don't model
execution, and most people using them don't know that.
if you use one, assume your results are optimistic,
more so on bigger size and younger tokens.

tx hash and mint in the replies if you want to check the
numbers yourself.

[Ferre @FerreWeb3]

Nah bro, a random coin you buy isn’t just going to do 1000x

Nothing is random

You need to read people, cut losses, understand the meta and know which ticker fits it best

That’s the difference between having an edge and just spinning slots

[totoonsol @totoonsol_]

your buy not moving the chart is the real issue, and that part is fixable (thats what im trying to do btw). psychology isn't

but early on it's still worth it. watching is passive, taking the trade makes you actually look at why this one ran and that one didn't. plus you learn the terminal and build reflexes without paying for it

[totoonsol @totoonsol_]

paper trade first → go live one you’re ready

[Ray @ray_world0]

learn to trade on ur own

[Ray @ray_world0]

learn to trade on ur own

[totoonsol @totoonsol_]

@BellounOFF @Luckyx1o0 paper trading is great, dryflip is not

[totoonsol @totoonsol_]

@dexfontaine tbh they’re a bit all the same, basically wrong.

i’m building one, free forever (i think it’s crazy some tools without naming them make you pay for paper trading), open source and i’ll give free sol to some users one everything is setup

[totoonsol @totoonsol_]

stop using dryflip

every paper trading tool i've looked at simulates the trader. the buttons, the chart, the pnl counter

none of them simulate the chain

your buy fills at the displayed price, instantly, always succeeds. that's not what happens when you click for real

i'm trying to build the other one

[totoonsol @totoonsol_]

first thing: check whether your txs are failing or reverting. two different problems

never landed = priority too low, or the bundle didn't get picked up.

landed and reverted = your slippage
tolerance, the curve moved past your limit before you arrived

0.01 max auto fee is fine on a quiet mint, low on anything actually being fought over. and MEV protection adds a landing condition, so in pvp you're trading fill rate for protection

no setting gives you 100% fills though

[totoonsol @totoonsol_]

paper trading teaches you a market that doesn't exist.

small PSA for anyone practising memecoins on a
simulator before putting real money in. the fills
you're seeing are not what you would have gotten on
chain, and i measured exactly how far off they are.

pumpfun's bonding curve is a constant product pool.
it starts at 30 virtual SOL and 1.073b virtual tokens.
two things happen on every buy that most paper models
skip entirely:

1. a 1.25% fee (0.95% protocol + 0.30% creator) is
taken before your SOL ever reaches the curve.

2. your own order moves the price while it executes.
you get T * x / (S + x) tokens, not x / (S/T). the
bigger your size relative to the pool, the worse your
average fill.

i implemented this from the program IDL and validated
it against 119 real trades across 3 tokens at different
points on the curve. tokens, net SOL and fees match to
the lamport on all 119. so the numbers below aren't a
model of a model, they're the chain.

worked example, real buy at slot 448,107,312:

1.782 SOL in
actual tx received 30,636,219.502858 tokens
my implementation 30,636,219.502858 tokens
displayed price minus nothing 32,314,949 tokens

that's 5.48% more tokens than the chain would give you.
1.25% is the fee, the rest is your own impact.

and it's not a constant offset, which is the part that
matters. on a fresh curve a 10 SOL buy overstates your
fill by about 26%. on a curve near migration the same
buy is off by about 10%. a 0.1 SOL buy is off by 1.5%.
so the error scales with exactly the behaviour new
traders are trying to learn: size up, enter early.

selling has the same problem in reverse. paper tools
show unrealised pnl at the displayed price, but exiting
a large position pushes the price down as you sell. the
number on your screen is not what hits your wallet.

why this matters: a lot of people practise on +5 to +15%
moves. a strategy that looks profitable on paper can be
flat or negative live purely from execution, before you
even count priority fees, jito tips, failed transactions
and account rent.

i'm not selling anything and i'm not naming any tool.
the ones out there are genuinely useful for practising
workflow and discipline. they just don't model
execution, and most people using them don't know that.
if you use one, assume your results are optimistic,
more so on bigger size and younger tokens.

tx hash and mint in the replies if you want to check the
numbers yourself.

[totoonsol @totoonsol_]

@dexfontaine none i'd trust for size. they all fill at the displayed price, which can't happen on a curve

what other one do you mean by hiccups?

i’m building one that reflects what the chain actually does to your order. fees, impact, latency, failed txs (for free ofc)

[totoonsol @totoonsol_]

if you're paper trading memecoins, what would make you feel ready to go live?

[totoonsol @totoonsol_]

tested a few of these against real on chain fills recently

the gap is the execution. on a bonding curve your own buy moves the price while it fills, so you can't get the number that was on screen when you clicked. most of them give it to you anyway. none of the ones i checked subtract the 1.25% pump takes each way

i rebuilt the curve math and matched it against 119 real trades, 119/119 to the lamport, so i can tell you how far off a given fill is if that's useful for the video

[totoonsol @totoonsol_]

what "i'll learn as i go" actually looks like
in memecoins

[totoonsol @totoonsol_]

how slippage works, short version

slippage is just the difference between the price you saw when you hit buy and the price you actually get filled at your tx is not instant.

it goes out, sits there, lands a bit later. sometimes a few hundred ms, sometimes worse while it’s sitting there other people buy. the curve moves. you get whatever’s left when you finally land, not what was on your screen

example: you send 1 sol into a new mint. at the moment you clicked that 1 sol was worth 34.6m tokens then 5 sol lands before you. now your same 1 sol only gets 25.5m

26% less tokens. same click. same size.

that’s just how it works when the thing is moving the slippage % in your settings is only how much of that you’re ok eating before the tx just fails instead of filling worse

that’s all it is

hope this helps

[totoonsol @totoonsol_]

yes ~1.7% at 0.15 SOL on a fresh curve. almost all of that is just the 1.25% fee, the impact part is only about 0.5% because you're tiny relative to the 30
virtual SOL. the curve barely notices you.

that's the whole point: impact scales with your size relative to the pool, so small size is genuinely protected from it.

16% to fees and 133 of 583 failing is exactly what none of these simulators model. a paper tool would have shown you 583 successful trades and no fee drag at all.

would you be up for sharing that wallet? measuring real failure rates is the next thing i'm working ! thx

[totoonsol @totoonsol_]

the virtual reserves only set the starting price, they can't be withdrawn. real SOL reserves start at zero, so nobody can take out more than actual buyers put in.

but your instinct is right for a different reason: a dev bundling at the bottom of the curve buys with real SOL at a price nobody else will ever get, and can dump that back onto later buyers. nothing in the program stops that.

[totoonsol @totoonsol_]

if you trade pumpfun and don't know what the virtual
reserves are, this is worth 60 seconds.

every new token starts with 30 virtual SOL and 1.073b
virtual tokens. nobody deposited them, the program
invents them so there's a price from second zero.

price is just the ratio: 30 / 1.073b.

you buy, SOL goes in, tokens come out, ratio moves.
that's it. that's the whole chart. the curve completes
at about 86 SOL in, fees included, and the token
migrates.

everything else people say about the "chart" is just
this equation and other people's orders.

[totoonsol @totoonsol_]

your buy is not a button. it's a transaction that has to compete for a spot in a block, and it can lose.

most people learning memecoins have never been told this, because their simulator has never said no to them.

[totoonsol @totoonsol_]

gm time to lock in

[totoonsol @totoonsol_]

@dexfontaine nice way to try new strategies keep going bro! which paper trading extension are you using ?

[totoonsol @totoonsol_]

crypto twitter talks about narratives. the trenches are the most measurable market in history.

every trade, every wallet, every launch is public and permanent. you can replay any minute and see exactly who bought what and when.

in equities that data costs six figures a year and you'd still see less.

and yet almost everyone here trades on vibes.