The approach, in plain language

Every BasisPoint tool is built from the codebase of a working personal trading bot that runs around the clock in paper-trading mode. It uses real market prices and simulated order fills, so its behaviour can be watched without real money at stake. A connection to a broker's paper-trading account is part of the design, and live trading is blocked in the code itself until the account owner explicitly enables it.

The bot looks at each instrument through several independent sources. Each gives a view from -1 (points down) to +1 (points up). A source with nothing to say gives no view rather than a guess.

Price momentum

How strongly the price has trended over one, three and six months, adjusted for how much it usually swings.

Technical indicators

Standard chart measures: 50- and 200-day moving averages, RSI, MACD, Bollinger bands and 20-day rate of change.

News (signals pack)

Fresh headlines about each company, read by AI. Only new facts about the business count - results, contracts, approvals, lawsuits - never price recaps, lists of stocks or opinion pieces.

Public trading disclosures (signals pack)

Official filings of stock buys by members of the US Congress. Sales don't count, and each member's buys are graded by how often their earlier buys beat the S&P 500. Shown with their record; they get no say in trades while that record doesn't earn it (today it doesn't).

The views are combined into one score using weights. Different holding periods lean on different evidence - short-term ideas weigh recent trend and news more, long-term ideas weigh the long trend.

Every source counts by its record. Two more read company figures - value (a company's earnings against its price, compared with others in the same part) and earnings surprises (the last report against what analysts expected) - and were tested on twenty years of S&P 500 companies. Each source's weight follows how often its views beat the S&P 500, live and over ten years of history, with a short record pulled towards a coin flip; a source no better than a coin flip gets no say until its record improves. The news and Congress signals were rebuilt in November 2026 and start a fresh record. Ideas that showed no edge at all in the same research - analyst rating changes, the yield curve, the dollar, market breadth, seasons - were left out.

Every suggested trade carries two numbers. Agreement is the share of all sources' weight that points the same way as the trade; sources that stay silent count against it, so a thin case can't look strong. The bot's rules (the minimum for a purchase, when a large trade must wait for you) use the agreement.

Chance is how often purchases like it actually did better than the S&P 500 fund on the bot's own checked record, after the fees of the trade. It is measured, not guessed: with a short record it stays near a coin flip, and it never goes below 10% or above 90%.

• Neither number ever reaches 100%. No signal is certain, and the tool never claims otherwise.
• When the record shows no edge after fees, the bot doesn't buy a single stock just to be busy: the money stays in the index fund, the best-evidenced option.
• Large trades (in the reference bot, 20% of the portfolio or more) with agreement below a minimum (60% by default) always wait for a person to confirm them.
• The bot keeps score: every day it notes what each source says about every stock it watches, and later checks whether that stock then did better or worse than the S&P 500 - over a week for short holds, up to half a year for long ones. The weights shift slowly, in small capped steps, and each change is logged with its reason.

Nothing happens silently. Every plan, trade, weight change and settings change is logged with the reasoning behind it, in plain language. A built-in assistant can answer "why did it do that?" from those logs - it can propose actions, but never carries them out without confirmation.

• Every term in the app - from "drawdown" to "rebalancing band" - is underlined with a plain-language definition one tap away, and a glossary lists them all.
• An optional weekly summary arrives every Friday: the week next to the S&P 500 fund, how far each part has drifted, and the trades made.
• Every optional feature has its own on/off switch in the settings.

Test a strategy on past prices before trusting it: the bot as it's set up, a mix you put together from a search box, or rules you describe in your own words. Any period from this quarter back as far as the data goes - up to 50 years - always next to simply holding an S&P 500 fund, with fees, currency costs and tax included.

• Every decision uses only what was known the trading day before, so the test can't peek at the future.
• Before a fund existed, the closest older series stands in for it - always labelled, never hidden.
• Seven checks ask whether a result is real or luck: years it never practised on, every start date, double the fees, trading a day late, crashes, a thousand reshuffled histories, and how many ideas you tried on the same period. The answer is one of four: Signal, Unclear (ahead, but ten years can't rule out luck), Mostly noise, or Behind the S&P 500 (simply holding the fund did better after the same costs).
• Before you switch, the bot compares the tested strategy with your bot as it runs now: the result a year, the worst fall, the luck verdict and what switching costs today.
• Backtests are hypothetical. They show how rules behaved in the past, never what they will do next.

• By default every plan waits for approval. Automatic paper trading is a switch you turn on yourself.
• Each part of the portfolio has a hard cap. Anything above it is trimmed and redirected, and the reason is logged.
• Positions have minimum holding times, so the tool doesn't churn on day-to-day noise.
• Small drifts are ignored, and trades whose fees would eat too much of their value are skipped. If a part drifts far from its target between check-ups, an optional drift alert tells you.
• Live trading is refused in code until the account owner explicitly allows it.
• Optional brakes - stop-losses, trailing stops, a circuit breaker and limits - start from Light or Firm presets, and before you switch a stop on the bot shows what it would have done on your own companies over ten years: how often it would have sold, and how often the price was higher again three months later.
• Every sale made by a rule says why with its own numbers, and where the money goes in the same plan.

• Public trading disclosures arrive up to 45 days late and only give amount ranges.
• Some free market-data sources are unofficial and can change or stop without notice.
• Projections are statistical scenarios from historical data, not predictions. Past patterns may not continue.
• No tool can guarantee results. There are no promised returns here - only a clearer view of your own decisions.

Your bot is a copy of ours, set up from a form of clear choices: which markets (the US, emerging markets, all-world and eleven more, each as one fund), which extra funds (bonds, property, cash-like), which themes of single stocks, the currency it reports in (eleven), how often it rebalances (daily to quarterly, or only when something has drifted), which signal sources it listens to, which protections it uses (stops, a circuit breaker, limits on position size and number of holdings), how decisions reach you, and what it is compared with.

Every bot can run up to two paper bots beside it: full BasisPoint bots on practice money, each with its own strategy, portfolio, plans, settings and messages - the same strategy and learned weights as your bot, or one designed from 0 with the blueprint's questions before it is created. You switch between them from the menu at the top left, and every Telegram message names the bot it is about. Only your main bot can trade real money. The optional Strategy Lab is the workshop for new strategies: a ready-made textbook strategy, or your own rules built in a step-by-step form (which sectors, markets or companies - found by name, how it chooses, stops and limits). Every strategy is tested over ten years and marked signal or noise; the ones that fit your own mix and hold up are recommended first, and when none holds up the bot says so and names the S&P 500 fund as the best-evidenced option. To watch one day by day, practise it beside your bot in the Vault or try it in a paper bot. Backtests are hypothetical. A very different approach from ours is a separate, quoted job.

The Vault keeps every version of your strategy with its own record, and up to three other strategies practise beside it on paper - replayed every evening on one chart - so you can see how a different mix or a Lab strategy would be doing before you switch. Repeated answers in the optional quiz become preferences for a stock or a whole part that the bot checks against the S&P 500 by itself every month, halving one when it turns out wrong.

Support commitment: Any bugs identified in your tool will be fixed at no additional charge, and every update is free.