I have spent about a year on this.
Some of it went into a WordPress plugin I eventually scrapped, some of it into research and planning that left no trace, and the rest into building what you can read today. Almost none of that time went into making it look like it works. Most of it went into stopping it from lying to me.
Building a trading platform while raising four children on my own is a decision I would defend in court and struggle to explain at a dinner party. But here we are.
It is open source now. The parts people normally pay for are free: the scoring engine, the bots, the backtesting. And I need to be straight with you about the reason I am writing this, because I need the project to earn something. That is an uncomfortable thing to type, and I would rather type it than pretend otherwise and quietly build a paywall later.
Where it came from
It started as something else: TradePress, a WordPress plugin. I scrapped it. It was a prototype more than a product, and the idea quietly outgrew the platform it was written on, which is a politer way of saying I built the wrong thing, learned a lot, and started again.
I wrote about that chapter at the time, in TradePress: Why I Built It — frustration, ambition, and three monitors. Much of what is in this post was already in that one as an intention. This is the part where I found out which parts were true.
So this is roughly a year of work, and not all of it is in the commit history. A great deal of it was research and planning, the kind of work that leaves no trace and looks like nothing happening if you are watching the repository.
The honest reason I built it. I am a solo parent of four children who live with me seven days a week. My working hours are limited, and so is my income. Trading appealed to me as a long-term income source and, separately, as a genuine interest. I enjoy chart analysis, and I got 70% on my first portfolio across my first two years of trading, which was enough to make me wonder, one day, about doing it properly.
That 70% is my own trading, on my own account. It is not a result from this software, and it is not a claim about anything it can do for you.
What I actually needed was not another strategy. It was a command suite that could watch several brokers in one place, so I could trade whichever kind of trading suited the time I had that day, without losing my scarce hours to clicking between terminals. That is still exactly what it is for.
The diary I didn’t keep
People tell you to keep a journal when you build something. I didn’t. I kept building instead, and somewhere around month three I realised I could not remember what I had done the week before, let alone what I had learned.
Then I checked the repository and found it had been keeping a diary for me. Nine hundred and forty-seven commits. Four hundred-odd closed issues. And, because I got tired of re-arguing the same decisions with myself, a register of 147 settled decisions — each one dated, each one with the reason it was settled, and a note on whether anything actually enforces it.
Reading that back is the closest thing I have to a diary, and it is more honest than a diary would have been, because it records the failures with numbers attached.
Three moments I’m glad are written down
1. The backtest that reported a 1,380% return
A backtest came back with $138,058 profit on a $10,000 account — a 1,380% return — over 610 trades. That number looked like a discovery. It was arithmetic.
The sizing model asked for a fractional position. Something upstream rounded it up to one whole unit, because you cannot trade part of a unit. Except that unit was an entire index contract. The simulation was buying something roughly eight times the account on every trade, and the “profit” was the leverage doing the work, not the strategy.
The fix was not clever. The rule is now: if a position is unaffordable, refuse it, and say so. That protocol now returns 7 signals, 7 refusals, 0 trades, $0 — which reads like failure and is actually the most honest output the system has produced.
That is the whole flavour of this project. The dangerous bug is never the one that crashes. It is the one that produces a confident, plausible number and lets you believe it.
2. The system was quietly reading tomorrow’s news
Every backtest was scored against the present. Historical bars were being judged using today’s fundamentals, because the code fetched the current value and never checked when it was published.
So a strategy “knowing” the result of an event it could not have known about looked brilliant. Every backtest of it was fiction. Nothing errored. Nothing warned. The number simply appeared, and it was wrong in the most expensive possible direction.
Both data sources that did this now have to prove their own timestamps, and if they cannot, the directive is not allowed to score. It reports “unmeasured” instead. My favourite outcome in this codebase is the one that says: I cannot judge this.
3. The trailing stop that placed nothing
For weeks I assumed my exits were losing money because the trailing stop was too tight. So I tested it — six variants against a baseline, on the same window, same protocol.
None of them helped. Every variant cut the average win roughly in half, and none beat the baseline. And while checking, I found something worse: the simulated trailing stop was not placing or moving anything at all. It had been enabled in my head and disabled in the configuration, on all forty bots.
Then, once the measurement was actually about exits rather than entries, the real number appeared: 42 of 95 closed trades had peaked in profit and closed at or below zero. Not a trailing-stop problem. A give-back problem, which is the price of letting winners run rather than a missing feature.
I would never have found that by reading the code. I found it by making the measurement capable of seeing the thing I was asking about.
It is actually out, and that is genuinely exciting
I want to put this before the money part, because it is the truer feeling.
For a year this thing has run on one machine, watched by one person, with me as its only critic. There is a particular kind of loneliness to building something nobody else can see. Every “does this make sense?” was answered by me, badly, at 1am.
The day the repository went public, someone could read it. Someone could disagree with a decision I wrote down. Someone could run it and find that the thing I was proud of confuses them immediately, which is useful, and which I had been denied for months.
That is the part I did not expect: open sourcing is not a launch, it is the first time the work is allowed to be judged. I am genuinely looking forward to being told what is wrong with it. Months of guessing whether the design holds up is worse than being corrected.
The part I would rather skip
Here is the honest position.
This project is licensed AGPL-3.0 and the source is public. That means I cannot really lock it. If I put a feature behind a paywall, someone can remove the gate. Not out of malice; that is just what open source means, legally and practically. So any money this earns has to come from people choosing to support it, because it is useful to them and they would like it to keep existing.
I would also like to be candid about why that matters: I need an income from this. Not a nice-to-have, not a lifestyle-business footnote. I have spent a year of my life on it, and the honest version of this post has to include the sentence — I am hoping this earns something.
What I am not going to do is make the useful part worse to manufacture a reason to charge.
The core stays free. Scoring, gating, bots, positions, risk controls, backtesting. The part that would actually cost you money if it were wrong stays free, permanently. Charging for risk controls would be indefensible.
If you would like to support it, there is a page that lists every option — GitHub Sponsors, Patreon, PayPal or Buy Me a Coffee — and none of them change what you can use.
The one thing that may become paid is the Lab: the advanced backtesting suite, with highly visual side-by-side A/B comparison of two configurations over the same bars, showing what changed and what it was worth. It is built for people who know what they are doing and would rather pay for a serious tool than be given a shallow one.
And I want to be specific about the end of that arrangement. When the income I have received after tax, from every source and including the project, reaches one million pounds, the Lab becomes free for everyone. That is not a marketing promise. It is the point at which I would no longer need it to be paid, and I would rather name it now than quietly forget it later.
Nothing that already exists will be taken away. If it is free when you read this, it stays free.
I would rather give away something genuinely valuable and ask for support than water it down and insist you pay for the good version. That is not a marketing position; it is a bet that quality is the only thing that actually travels.
What I would like from you
None of these cost money:
- Try it on a demo account. Not a funded one. It is pre-release, and it is not configured to make money for you. Please read the disclaimer before you connect an account.
- Tell me when a page confuses you. An outsider’s confusion is worth more than my own review.
- Star the repository if you want to find it again. It is, embarrassingly, the main signal I have for whether anyone cares.
- Tell me what would make it worth supporting financially. Genuinely. I would rather build the thing people want to fund than guess.
Looking forward to hearing if that is useful to you. If it is not, thank you for reading this far.
— Ryan



