Score the market.
Track your edge.
Bubble and systemic risk on any market, a trade journal that prices every instrument on its own contract, and macro events ranked by impact. BubbleGuard and JournalGuard are live today, on an engine replayed over past crises before it scored anything live.
Three modules, one account
Market risk, your own track record and the macro calendar, adapted to the way you trade. Free for every registered user.
BubbleGuard
Bubble detection and systemic risk: one 0-100 score per market, from nine independent quantitative modules.
Open BubbleGuard →JournalGuard
Trade journal that prices futures, CFDs, forex and crypto on their own contracts. Manual entry today; calendar, statistics and prop firm limits are being built.
Open JournalGuard →NewsGuard
Coming soonMacro events and market news, ranked by expected impact across timeframes, filtered on the markets you trade.
Multi-Layer Bubble Detection
Six decades of academic research condensed into a single actionable score updated in real time.
LPPLS Model
The same model (by physicist Didier Sornette) that spots the accelerating, super-exponential price rise that comes right before a crash.
GSADF Testing
Detects when a market enters an explosive, runaway regime — the statistical fingerprint of a forming bubble.
9-Module Score
One simple 0-100 score blends 9 signals — price, valuation, macro, credit stress, sentiment, and machine learning — into a clear verdict.
Real-Time Alerts
Get notified the moment a market you follow crosses into dangerous territory. Set it once, stay protected.
Risk Management
Every score comes with historical context — see how today compares to past bubbles, so you know how worried to be.
Built on Real Data
Replayed over labelled historical crises with point-in-time data: the score crossed its alert level before the peak in eight of the nine episodes that could be tested, 19.8 months early on average — a read on the regime, far too early to be an exit.
Replayed over past crises
Retrospective testThe engine was run backwards over labelled historical episodes, one quarterly date at a time, seeing only the prices that existed on each date. Four modules can be reconstructed that far back — LPPLS, GSADF, technical and credit stress. Valuation, sentiment and microstructure have no history at this granularity, and the machine-learning module was trained on these same episodes, so it was left out rather than allowed to grade its own homework.
8 of 9
Flagged before the peak
The remaining one, cannabis stock bubble 2018, had no testable point: TLRY had been listed two months at the peak.
19.8
Months early, on average
Early enough to be useless as an exit — which is the whole point below.
~1 in 4
Calm readings above the alert level
Measured on 25 quarterly readings from calm markets (S&P 500 and Nasdaq, quarterly, 2010-2017).
What 21 months of warning is, and what it is not
The 2007–09 crisis was flagged in January 2006. The S&P 500 peaked in October 2007, 21 months later. The dot-com flag came 29 months early, in October 1997, and the Nasdaq went on rising until March 2000: anyone who sold on that signal would have sat out years of the largest part of the rise.
So this is not a timing tool and not an exit signal. It reads the regime you are trading in — how much of the rise is running on itself, and how much systemic risk has built up underneath it. A high score is a reason to look, not a forecast: to size differently, to know what you are holding, to stop assuming the last two years describe the next two.
In the test
Read this for what it is: a test run after the fact, on crises that were already labelled as crises. No score in it was published while those markets were rising, and a model that fits the past is not the same thing as a model that will call the future. How the score is built.
From Data to Insight in Seconds
Our pipeline ingests global market data, runs 9 independent models, and delivers a single composite score.
Ingest
Real-time price, volume, sentiment, and macro data across global equities, indices, and crypto.
Analyze
9 quantitative modules run in parallel: LPPLS, GSADF, valuation, credit stress, ML signals.
Alert
Composite score published < 5 min. Alerts fire automatically at your thresholds.
Your track record, counted correctly
A journal is only as good as its arithmetic. Futures, CFDs, forex and crypto turn the same price move into different amounts of money, so JournalGuard prices every trade on its own contract instead of adding up points.
One Nasdaq trade, ten points in your favour, three instruments:
$200
40 ticks at $5 each
$20
same index, a tenth of the size
the broker's lot
asked once, never guessed
Adding those points together into one number would invent money. That is also why a CFD account will not price a trade until you say what one lot is worth: two brokers disagree, and a plausible wrong figure is worse than a question.
Available today
LiveLog a trade in one form
Instrument, side, size, entry, exit, stop, fees, setup, tags and notes. Nothing is computed from a guess.
P&L with the right multiplier
Futures priced in ticks, forex in standard lots, crypto per coin, CFDs at the contract size your broker uses.
Accounts and instruments
25 instruments ready out of the box, and a CFD account that refuses to price a trade until its contract size is set.
Being built
Not available yetCalendar dashboard
Every day of the month with what it made or cost, so a bad week is visible before it becomes a bad month.
Statistics that mean something
Expectancy, distribution of R multiples, drawdown — built on the P&L above, which is why that comes first.
Prop firm mode
Daily loss limit and maximum drawdown tracked against the rules of your account, while the day is still running.
File importers
Broker statements read in instead of retyped. Later than the rest: manual entry has to be right first.
Listed so you know what the module is for, not as a promise with a date on it. Open JournalGuard today and you get the entry form and the trade list: nothing on this side of the page is in there yet.
Not another news feed
A feed tells you an event happened. NewsGuard rates what it is likely to do to price, by asking two separate questions and keeping the answers apart.
How big is the event
A central bank decision and a second-tier survey are not the same size, whatever the headline says.
How much was already priced in
The market spends weeks pricing what it expects. What is left to move price is the part nobody had.
Neither question is worth much alone. Together they separate the four things a headline can be:
Major, expected
Moves little
The number everyone forecast correctly. It was in the price days ago.
Major, unexpected
Moves markets
The event that matters and lands where nobody was positioned.
Minor, expected
Background
Worth knowing it happened. Not worth changing anything for.
Minor, unexpected
Can still hurt
Small on paper, and still enough to run stops when no one saw it coming.
The most important release of the month can be the one that moves price least, and a secondary figure nobody was watching can be the one that runs the day. Importance alone cannot tell you which.
Rated per timeframe
One rating for a scalper and a swing trader would be wrong for both, so each event is scored at four horizons:
5 min
The spike: who gets caught in the first print.
1 hour
Whether the move held once the desks read it.
1 day
What the session closes as, after the reaction.
1 week
Whether it changed the trend or was only noise.
The day and week ratings are calibrated on price history we already have. The five-minute and one-hour ones cannot be: reconstructing them for past events would need intraday bars nobody keeps for free, so they are measured forward from the events that happen after launch — and until there are enough of them, they will say so instead of showing a number.
On your markets
Filtered on what you actually trade, taken from your profile. An index trader has no use for a coffee inventory report, and it should not be on the screen at all.
NewsGuard is not built yet. This is the idea it will be built on, so you can tell us now if it is the wrong one.
Everything you need before you press the button.
Market risk, your own track record and the events that move price — free for every registered user, no credit card required.