Scientific foundation

The burn is an estimate of how many grams of carbohydrate your body oxidises per hour. It is worked out stretch by stretch: each 30-second stretch contributes its energy (mechanical work divided by a reference gross efficiency of 20%, the usual figure in amateur and trained cyclists) times the share of that energy that comes from carbohydrate at that %FTP.

The carbohydrate share rises with intensity (Romijn et al., 1993). RATIO starts from what van Loon et al. (2001) measured in fasted trained cyclists and maps it onto %FTP assuming FTP is 73% of maximal aerobic power, an approximate conversion; below and above the measured points the values are RATIO's own extrapolation. Eating during the ride keeps that share up: with 30 g/h or more, 5 points are added (Newell et al., 2018; King et al., 2019; Rothschild et al., 2022); without eating, or on a ride under 75 min, the fasted curve applies. Stretches without pedalling count as rest. The metabolic phenotype (diesel, explosive) no longer moves this figure: there is no evidence that a label shifts the share by a fixed amount.

A real analysis uses Strava's second-by-second power. A plan, which has no such data yet, you pedal at your zone's power and, on a route, freewheel where its duration clock has you descending without pedalling (“How long you take”); without a route and on a trainer, at that power all the time. A real analysis without a power meter uses the estimated intensity, steady. Energy comes from the average power, which is the work you do, not from the NP. Because efficiency varies between people (18⁠–⁠22%) and the share is approximate (±10 points), the burn is always shown with its range. The UCI position statement notes that power meters allow energy expenditure to be estimated with reasonable validity.

What you burn and what you absorb. The burn is almost always higher than what we recommend you take: the gut cannot absorb everything you burn, and glycogen and fat cover the difference. So the burn does not set the ride target; it feeds the deficit and the recovery.

The kcal. The plan's kcal come from the same stretches as the carbohydrate burn, at the same efficiency (20%): on a route, the time you freewheel downhill and the time you brake at the top speed come from the same clock as the duration, and the rest is pedalled at your power. Structured workouts add up their own steps. A real analysis starts from the ride's work — the kJ Strava gives or, without them, the average power over the moving time, the estimated power or your intensity over your FTP — at that same efficiency (18⁠–⁠22%), and always shows it as a range. The calories Strava gives are not used.

References: Romijn et al. 1993; van Loon et al. 2001; Newell et al. 2018; King et al. 2019; Rothschild et al. 2022; Coyle et al. 1992; Jeukendrup & Wallis 2005; Burke et al. 2026

The target follows the duration-based guidance of the UCI position statement, Thomas et al. (2016) and Jeukendrup (2014): the longer the ride, the more carbohydrate per hour. RATIO uses three ranges:

Under 75 min you don't need any carbs; rinsing your mouth with a sugary drink is optional. In the other two ranges, the ride's intensity decides where the target sits, measured as NP/FTP (normalized power over your FTP): near the bottom of the range at an easy pace, near the top close to threshold. It is rounded to multiples of 5.

Your carb ceiling then applies. It starts from what you say you tolerate in Gut tolerance, in grams per hour (under 40, 40 to 60, 60 to 80 or over 80), with the top of each range as the ceiling (40, 60, 80 and 90 g/h), and then moves with the rides you log in “Your ride”: “How RATIO learns your ceiling” tells how. If you don't know, the target stays at the bottom of the ride's range, and the plan says so. If you mark the ride as a key ride and it lasts more than 2.5 h, the target is 90 g/h, to train the gut at race rates. Above 90 g/h (only with a target you set yourself) it is treated as experimental: the position statement notes that professionals report 120 g/h or more, but the benefit still needs validating.

Difference from the position statement. For rides over 2.5 h it recommends 90 g/h in racing. RATIO uses 60⁠–⁠90 g/h by intensity, because most of an amateur's long rides are endurance rides, and keeps 90 g/h for key rides.

The same intensity in the plan and the analysis. A real ride with a power meter is analysed with the NP Strava gives over your FTP (average power when there is no NP), because the average falls short on rides with freewheeling descents. Without a power meter it uses, in this order, heart rate, a power estimated from speed and climbing, or the effort you give. The planner's zones stand for these NP/FTP values, from Coggan's power levels (Allen and Coggan, 2010): Recovery inside its level (under 55%), Endurance and Tempo at the middle of theirs (55⁠–⁠75% and 76⁠–⁠90%); Threshold and VO2max can't be held for a whole ride, so a ride of those intervals takes the intensity factor Allen and Coggan give an interval session with its recoveries (0.85⁠–⁠0.95), and a race the boundary between a long one (0.85⁠–⁠0.95) and a short one (0.95⁠–⁠1.05). The same on a trainer and for the duration:

Ride type. You choose whether the ride is steady, a group ride on the wheel, or a chaingang with attacks. For now all three use the same steady-ride template: the shape of group power comes from a single rider and needs validating with more. The only thing a chaingang changes is the NP/FTP: 0.90, that of an interval session with its recoveries (Allen and Coggan, 2010): the pulls are the work and the wheel the rest. It holds whatever zone you pick (the zone only estimates the duration). In the two group types, in Advanced mode, you can enter the easy minutes out and back: those minutes are computed at the Recovery NP/FTP (0.50), the rest at the type's, and the target uses the NP of the whole ride. On a chaingang the target is still the one its duration and intensity give, as for any ride (the consensus doesn't raise it for irregularity); if it is below 60 g/h, a note says you can get close to that figure if you tolerate it. It is a theoretical proposal (Morton et al., 2026) backed by a single trial, not a recommendation.

Low-carb training (Train Low). Only when you choose it, in Advanced mode and in Recovery or Endurance: you set off without a carbohydrate breakfast and take no carbohydrate the first 2 hours (a mouth rinse, if you like); if the ride is longer, from there the target is 30 to 60 g/h, placed by your intensity like the other ranges and never above your ceiling (Morton et al., 2026, table 2; in Impey et al., 2016 and 2018, the low arm took 0 g). It is a one-off strategy for specific sessions: only if the planned intensity doesn't suffer, and the position statement points out that evidence of performance benefits is still limited.

References: Burke et al. 2026; Thomas et al. 2016; Jeukendrup 2014; Jeukendrup et al. 2026; Morton et al. 2026; Impey et al. 2016; Impey et al. 2018

The gut absorbs glucose (and maltodextrin) through a transporter that saturates at around 60 g/h; fructose uses a different one. So above about 60 g/h you need glucose and fructose mixes. RATIO adjusts the home recipe's ratio to the target rate:

Above 90 g/h only a target you set yourself gets through, in Advanced mode and with the RATIO 1:0.8 mix in the bottles: it takes a mix with at least 35% fructose (1:0.8 gives 44%): with the 2:1 (33%), the planner limits the target to 90 g/h. Commercial products. Above 60 g/h, if what you carry with glucose and fructose doesn't cover at least 50% of what passes 60 g/h, the planner tells you and names the products with glucose only; for those it doesn't know (European labels only give “sugars”), it asks you to check the label. It never blocks (Burke et al., 2026). A target you set by hand goes up to 120 g/h; above 90 g/h it is marked experimental. The gut can be trained: taking increasing amounts in training improves absorption and tolerance (Jeukendrup, 2017; Costa et al., 2017); how RATIO uses it for your ceiling is in “How RATIO learns your ceiling”.

Bottle concentration. The recipe stays at about 8% carbohydrate (44 g per 550 ml), with no warning. In extreme cold (under 8 °C) the concentration drops to 60% and the rest of the carbs move to solids or gels. Each bottle shows its real concentration against 8%, the UCI's (Cheung et al., 2026) and the ACSM's (Sawka et al., 2007) reference for a drink: going past it is not an error — many drinks for high intakes do — it may leave the stomach a little more slowly, and a trained gut tolerates it better. Our own criterion, not a source's: above 12%, or 10% with glucose only, it may sit worse (dissolve it in more water); in the lab, glucose and fructose mixes of 10 to 15% have been tolerated (Podlogar et al., 2022; Pfeiffer et al., 2010).

What goes in each bottle and the flask. You set it: the product in its unit (grams of a powder or sugar, sachets or units, ml of a ready drink) and the ml of water. RATIO recommends an amount in steps you can measure and counts the one you set. The ratios behind “Fits” come from the review by Rowlands et al. (2015), where 0.7:1 to 1:1 fructose per glucose was absorbed fastest and 0.8:1 gave the most energy at 1.5⁠–⁠1.8 g/min, and sucrose from Trommelen et al. (2017): taken with glucose, it is oxidised as well as glucose with fructose. The product takes up room: each gram of dissolved sugar or maltodextrin adds about 0.62 ml (sucrose's apparent specific volume in water, ~0.63 ml/g; Tinjacá et al., 2018), so 500 g with 280 ml of water measure about 590 ml. The water starts by filling to the top (the capacity minus the product, in steps of 10 ml); if water and product don't fit, it says so with the figures and limits nothing. The concentration is measured over the final volume, water plus product, and the liquid that counts is the water plus a ready drink's ml. A bottle left unset counts for nothing. Table sugar: 100 g of carbohydrate per 100 g, glucose and fructose 1:1, no sodium (USDA FDC 169655).

References: Jeukendrup 2010; Jentjens & Jeukendrup 2005; Rowlands et al. 2015; Trommelen et al. 2017; Jeukendrup 2017; Costa et al. 2017; Burke et al. 2026; Tinjacá et al. 2018; Cheung et al. 2026; Sawka et al. 2007; Podlogar et al. 2022; Pfeiffer et al. 2010

Your carb ceiling is the most the plan asks you to take per hour, even when the ride would call for more. It starts from your answer in Gut tolerance (40, 60, 80 or 90 g/h, depending on the range you pick) and then the rides you log in “Your ride” adjust it: how much you took, how long it lasted and “How did you feel?”. Only rides of 2 h or more that you log after your answer, and on which you took carbs, count: without them there's nothing to learn about how many you tolerate. New plans use your ceiling at the time; saved ones don't change.

Learned, or from your answer. As soon as a ride moves your ceiling, step 03, Profile and “Your plan” say it is learned from your rides, even if it later comes back to your answer's figure, as in the example. While none has moved it, they say it comes from your answer in Gut tolerance. If you change your answer, or pick one in Profile instead of your learned ceiling, the count starts again from it: earlier rides no longer move it. If your answer was “I don't know”, which is no longer offered, it becomes a figure of your own with the first long ride on which you have stomach trouble taking 40 g/h or more, or take about 60 g/h or more without it.

What the evidence says. The gut trains: taking carbs again and again while you train lets you use more of what you take and makes it easier on your stomach. In cyclists who took carbs in every session for 28 days, burned more of the glucose they drank while riding (Cox et al., 2010); in runners who took carbs while running for two weeks, stomach trouble fell by 60% (Costa et al., 2017a). Real gut trouble shows from about 2 h of exercise (Costa et al., 2017b): that's why only those rides count. And it rises with what you take (Pfeiffer et al., 2012): that's why it goes down with stomach trouble. The floor of 30 g/h is the least the guidelines ask for on a ride over 75 min (Jeukendrup, 2014; Thomas et al., 2016), and above 90 g/h an intake counts as experimental (Burke et al., 2026).

It's a rule of our own, a prudent one: no study has tested how many rides it takes or how much to go up or down each time; the trials use a fixed dose or let each rider build up over a few weeks. We chose steps of 10 g/h, because a bigger jump on the next long ride is the very risk of stomach trouble this avoids; 2 rides near your ceiling before going up, and just one with stomach trouble to go down, if you took at least 40 g/h, or your ceiling minus 20 if that is less; and a margin of 5 g/h, what the plan's targets are rounded to. Nor does it expire: nobody knows how long the training lasts. You see it in Profile, under “What your rides have taught RATIO”, and change it in Gut tolerance.

References: Cox et al. 2010; Costa et al. 2017; Costa et al. 2017; Pfeiffer et al. 2012; Jeukendrup 2014; Thomas et al. 2016; Burke et al. 2026

The UCI position statement suggests keeping net fluid loss below 2% of body mass. You don't need to drink everything you sweat: RATIO gives you a range.

Bottles, refills and the timeline are sized on the minimum. On rides under 75 min, the minimum is drinking to thirst. On a route, RATIO looks at the temperature hour by hour, at the point of the course where you will be at that hour: what has to be drunk over the day (what you sweat minus the limit) is spread over the hours in proportion to what you sweat in each, so cool hours ask for less, but never nothing.

It is “to thirst” when the ride's mean is below 15 °C, or when what you sweat over the whole ride stays within that limit: you needn't drink to stay inside it. Without a route, the ride's temperature decides for the whole day. Any minimum under 150 ml/h is shown as “to thirst”. Unless your sweat is measured, the minimum never goes above 80% of estimated sweat (if it is, that limit doesn't apply): drinking more than you lose is the main cause of hyponatraemia, and the estimate can run high. On rides over 4 h the plan always says: “Don’t drink more than you sweat: if you weigh more at the end than at the start, you drank too much”; on shorter ones, only if it asked you to drink more than you sweat. When a maximum keeps the ride from meeting the day’s limit and the loss is over 2%, the plan says how much weight it is normal to lose and that drinking more won’t make up for it. Above 32 °C the position statements suggest losing no more than 1%; RATIO applies it only with your measured sweat (below), because with an estimate it would push you to overdrink. 1,000 ml/h is a practical limit: above it, the stomach often can't empty in time; if you sweat more, the planner warns you about the deficit to replace when you finish. Mix bottles and their reloads never go above the maximum either: if the ride's carbs don't fit in that volume at the recipe's concentration, the planner tells you how many grams are short, for you to pick from your pantry.

How sweat is estimated. It starts from your sweat level (the chart shows what each sweats) and adjusts it for heat (+2% per degree above 18 °C, with no limit; it falls in cool weather, down to 0.7 times the level), humidity (+0.4% per point above 50%, RATIO's own estimate), indoor riding (+18%, no airflow from riding; RATIO's own estimate) and power: sweat follows the heat your muscles produce (Gagnon et al., 2013), so more watts means more sweat. The power is your ride's average or, in a plan, that of the zone you pick; without an FTP, what your flat-road average takes, and without either, 150 W. It is an estimate, unless you measure your sweat.

Your measured sweat. In your Profile, “Measure my sweat”: weigh yourself naked and dry before and after a ride of 90 min or more, and note what you drink and pee (Baker, 2017). Each test is brought to 18 °C and 150 W by dividing by the same heat, humidity, indoor and power factors as above, so a hot test and a cool one can be compared. A test giving more than 3 L/h or less than 0.2 L/h is rejected: it is almost always a weighing error. With 2 tests or more, their median replaces your sweat level in your plans and in your ride analyses, and is shown with its margin: the error of two weighings on a home scale, about 150 ml, over what you lost in the test, never under ±10%. So a short test has a wider margin, and one past ±25% (under 600 ml lost) is saved but doesn't count; with just one, it is shown next to your level without replacing it, because the same person varies from day to day. The temperature is Open-Meteo's for the test's hours (indoors, the one you enter) and the power is your plan's, the one you type or that of the zone you pick.

References: Burke et al. 2026; Sawka et al. 2007; Gagnon et al. 2013; Cheung et al. 2026; Baker 2017; Hew-Butler et al. 2015

Sodium goes in what you drink: it is a concentration in the drink, not a share of what you sweat. What you need depends on how much of your sweat you replace and how much sodium your sweat carries (McCubbin, 2023). RATIO uses 600 mg per litre, in line with the 0.5⁠–⁠0.7 g/L suggested in the heat (Racinais et al., 2015); 900 mg/L if you are a “salty sweater”, about 40 mmol/L (Cheung et al., 2026); and 20% less if you have trained in the heat on 7 or more days in the last 2 weeks, because acclimatization lowers sweat sodium (Cheung et al., 2026).

There is a sodium target only if the ride lasts more than 2 h, it is 25 °C or warmer, estimated sweat reaches 1,200 ml/h, or you are a salty sweater and the ride lasts 75 min or more (Thomas et al., 2016). Otherwise it is 0: what you lose is replaced with the meals afterwards. The planner shows it as mg per litre you drink, with its equivalent in mg/h at the planned drinking pace.

The planner warns when a bottle goes above 1,500 mg of sodium per litre and suggests reducing the electrolytes or dissolving them in more water. The main risk of hyponatraemia is drinking more than you lose. Sodium only slows the fall in blood sodium when you drink roughly what you sweat over long efforts, and it does not prevent it if you overdrink (Hew-Butler et al., 2015). After the ride, if you log 2 or more times the target sodium, you are told: it is our own threshold, not a physiological one, so that you check how much salt the mix or the capsules carried.

References: McCubbin 2023; Racinais et al. 2015; Cheung et al. 2026; Thomas et al. 2016; Hew-Butler et al. 2015

Pre-ride meal (1⁠–⁠4 h before). The position statement recommends 1⁠–⁠4 g of carbohydrate per kg of body mass in the hours before, depending on what the ride will demand. RATIO sets it by duration and by whether it is a key ride, not by zone: a long, easy ride needs a tank as full as a hard one of the same length.

Carb loading. Only before a key ride over 2.5 h: 10⁠–⁠12 g/kg a day for 1⁠–⁠2 days (UCI position statement). It is shown as “Day −1 target”, with the dinner listed separately (~35% of the day, at most 3.5 g/kg). No loading is recommended for training rides.

What you drink before. The pre-load also includes water, between 5 and 7 ml/kg by your sweat on that ride (Sawka et al., 2007), and sodium only when the ride calls for it, at your plan's concentration. Caffeine. The dose, its limits and the sleep rule have a card of their own: “Caffeine and sleep”.

RATIO Plate. For the night before, the pre-ride meal and the meal after, RATIO also shows a plate level, the same figure told another way: “base”, a third of the plate is carbs; “endurance”, half; “load”, half with extras around it, only the night before a carb load. Before the ride it is “endurance” from 2 g/kg; the night before and after the ride, when the ride is longer than 2.5 h, and after the ride also when you train hard again within 8 h. The proportions are those of the Athlete's Plates of the US Olympic & Paralympic Committee and the University of Colorado, validated for a full day put together by dietitians (Reguant-Closa et al., 2019); applying them to a single meal, the cut-offs and the examples are our own criteria. It only shows if you turn on “Show food suggestions” in your profile; a new account has it off.

References: Burke et al. 2026; Thomas et al. 2016; Burke et al. 2011; Sawka et al. 2007; Reguant-Closa et al. 2019

Caffeine only goes into the plan if you ask for it (“Caffeine in this plan”, in step 02: off the first time, then as in your last plan) and with your bedtime: without it there is no dose. The dose is 3 mg per kg of your body mass (the ISSN gives 3 to 6; Guest et al., 2021), with two limits:

When. On rides of 1.5 h or more it is taken on the bike, not before the start (Whitfield et al., 2026); where it goes is in “On the bike: when to take what”. On shorter ones, 1 h before the start. A second caffeinated intake only on rides over 4.5 h, where the first stops covering.

The sleep rule. It comes from the meta-analysis by Gardiner et al. (2023) of 24 studies: to avoid losing sleep, about 107 mg at least 8.8 h before bedtime and about 217 mg at least 13.2 h before; we round to the prudent side. There are two other sources: the UCI consensus asks for 12 h above 100 mg (Whitfield et al., 2026), and a trial in 23 young men saw no effect on sleep with 100 mg 4 h before bed (Gardiner et al., 2025). We go with the most prudent. It is measured on what you have taken in the day by that hour, not on each intake alone. If the full dose doesn't fit but 100 mg does, it stays at 100 mg; if none fits, no caffeine today. Step 02 draws it under “Caffeine and sleep”: the most one intake may hold, hour by hour, until you go to bed. It is engine v19; until v18 they were 12 h and 6 h, and a plan saved then keeps its own.

What counts as caffeine on the bike. Everything you carry with caffeine, with the mg its brand declares: the gels and bars in your pocket, what goes in the bottles, the flask with its refills, and the sachets and drinks at the stops, a cola bought there included. It all adds up with the dose for the day's limits, and is checked at its minute against your bedtime. If something in your pocket breaks a limit, it leaves the plan and you're told which and why; a drink that breaks one says so on its row. Without your bedtime, what you carry with caffeine is checked against the day's limits only. What you take outside the plan, like your breakfast coffee, RATIO doesn't count.

References: Guest et al. 2021; Whitfield et al. 2026; Gardiner et al. 2023; Gardiner et al. 2025

The bottles carry the mix only up to what you will drink (the fluid minimum): that is what they give you if you drink as planned. They are counted stretch by stretch, from stop to stop: the most concentrated bottle is drunk first; every stop, water-only or with food, refills them with water and with the refill sachets you carry (one per bottle); with no stops, only the ones that fit on the bike count, until they run out; a bottle giving less than 10 g/h (what was left, diluted after a water stop) counts as water. A flask (a gel or one of your mixes, with its grams) is drunk at its full rate, what the target asks for after the bottles, until it is empty: one sip every 20 min (every 15 if a sip would pass 30 g), then the pocket takes over; a refill sachet tops it up at the stop where it would no longer reach the next one. Each sip goes with the water from your bottles that takes it to 12% carbohydrate (an approximation from studies with drinks; down to 8% is absorbed better), or the fluid minimum's if more; that water counts for when to refill, and with the bottles empty the flask waits. If over the whole ride the bottles' water and the flask's own pass what you sweat by more than 0.5% of your weight, the plan says so. What each stretch still needs for the target after the bottles, the flask and the refills comes from food, and each intake goes by its grams: the first at minute 20, whenever you last ate (if what you drink covers the target from the start, when it stops covering it), and each of the next when what you have eaten stops covering what the stretch asks for. The gap to the next one is that product's grams over that pace: a 50 g bar lasts twice as long as a 25 g gel, and if your bottles go from water to mix at a stop, the pace changes from there. Never less than 20 minutes after the previous one nor in the last 20 min, and another unit only if more than 5% of it is still uncovered. Your products take turns, one unit of each per round, solids first in each round. What the target doesn't ask for is a spare. If what you load falls short, it covers the ride from the start and you are told from when and how much is missing. No stretch goes more than 60 minutes without carbs where only the pocket gives them (the bottles hold water or have run out and there is no flask): if what you have left is only enough for that, the intakes go at that limit to the cutoff, and if you are short you are told how many are missing. Then they are placed on the route profile, and the timeline, the profile, the top-tube note and the exported route all use those same positions:

Caffeine. It goes before the hard riding still ahead, not before the first climb. Of the climbs starting from minute 72 and before the final cutoff, the one that leaves the most climbing in the 3 h after its base, the climbs behind it included, is chosen (a tie goes to the later one), and caffeine goes about 52 minutes before that base. Those 3 h of effect are a simplification of ours, not a measured figure: caffeine's half-life runs from 2.5 to 10 h (Guest, 2021). With no such climb, around 40% of the way, but never so early that its effect ends before the final cutoff. It never goes in the 52 minutes before the cutoff. A second caffeinated intake only exists where the first stops covering, about 3 h later, and if that leaves at least 60 minutes uncovered before the cutoff; otherwise there is no second intake. It stays where it is if it can; to make room for another intake, or because of the terrain, it moves 8 minutes at most. The other intakes are spread around it, starting from the first one, with similar gaps: none drifts more than 15% from the mean gap. Every caffeinated product you load is checked at its minute: if what you have taken so far would leave you too close to bedtime (the total, not each product on its own), if it would be a second intake on a ride of 4.5 h or less or has no place of its own, or if the day would go over 400 mg or 6 mg per kilo of body weight, it leaves the plan and you are told which one and why. The time of each intake. Riding minutes decide where each intake goes; the clock also adds the stops before it: 10 min for water-only stops, 15 for full refills, and the long stop if you enter one (the planned stop closest to mid-ride, or mid-ride if there is none). That time is what the caffeine margin before bed and the forecast hours are measured with.

Chaingang with easy minutes. If you enter a chaingang's easy minutes out and back, intakes go by phase: the first at the end of the easy start (never before minute 10); in the block, drinks only and one gel at most, “in the next sheltered or easy stretch, ±10 min”, and nothing solid (a caffeinated gel that falls in the block is that gel); the next at the start of the easy return, even inside the last 20 min. What doesn't fit in those three moments is eaten before the start, with the pre-ride meal. No study compares eating by phase; Jeukendrup (2026) advises finding the stretches of the ride that allow eating.

References: Pfeiffer et al. 2012; Jeukendrup et al. 2026; Guest et al. 2021

Carbs and protein. It depends on whether you train hard again within 8 h (the question in step 05, “no” by default). If not, a normal meal with carbohydrate and 0.3⁠–⁠0.4 g/kg of protein within the next 2 h is enough; you are also told how much more carbohydrate you burned than you took: what the model says you burned, minus what you log or, if you log nothing, what the plan carried. If so, the position statement's rapid refuelling: 1⁠–⁠1.2 g/kg/h of carbohydrate for 4 h, with 0.3⁠–⁠0.4 g/kg of protein.

Fluid and sodium. What you are short is what you sweated minus what you drank: what you log or, if you log nothing, the plan's minimum. If it is above 2% of your body mass, you replace 125⁠–⁠150% of it over the next 4⁠–⁠6 h, with 50⁠–⁠80 mmol/L of sodium (about 1,150⁠–⁠1,840 mg per litre) or salty food, as the position statement recommends. If not, drinking with your meals and to thirst is enough.

With rapid refuelling you can choose “Sensitive stomach”: 0.8 g/kg/h of carbohydrate with 0.4 g/kg of protein. Below 1.2 g/kg/h of carbohydrate, adding protein helps restore glycogen. There is no fat target: no source gives one.

References: Burke et al. 2026; Thomas et al. 2016; Beelen et al. 2010; Sawka et al. 2007

You define a mix the way you measure it at home: a batch of ingredients in grams and how many grams of that batch go in each bottle. RATIO scales the batch to that dose and adds up what each ingredient brings.

Glucose and fructose. Maltodextrin and dextrose count as glucose; table sugar as half glucose, half fructose. The carbs of a flavouring or a Pantry product have no split: they add to the carbs but not to the ratio. Energy is 4 kcal per gram of carbohydrate (EU Regulation 1169/2011); a Pantry product brings its own kcal.

Notes inform, they don't block. RATIO tells you, without stopping you from saving or using the mix, if the concentration is above 8% (a soft note) or 12% (10% with no fructose; a strong one), if it has more fructose than glucose, if sodium is above 1,500 mg/L, or if it has almost no sodium (under 100 mg per bottle) on a ride of 2 h or more or at 25 °C or more. Concentration always uses your bottle, not a water amount stored with the mix: your profile's in the editor and the plan's in the planner. Your mix's dose is never changed.

Where the base ingredients come from (per 100 g). Generic USDA FoodData Central values: table sugar (100 g carbs, 1 mg sodium; FDC 169655), fructose (100 g, 12 mg; FDC 169896) and salt (38,800 mg sodium; FDC 173468). Typical commercial label values, because USDA has no generic entry: maltodextrin (93.3 g; FDC 2423905) and dextrose (90 g; FDC 2412809); your brand may differ by a few grams. Sodium citrate (trisodium, dihydrate) is calculated from its chemical formula, C₆H₅Na₃O₇·2H₂O: 3 × 22.99 ÷ 294.10 = 23.45% sodium. Pantry products use the values you saved.

When you don't type the duration, RATIO estimates it with bike physics: the model of Martin et al. (1998), validated on the road (the power it predicts matched the measured one with a standard error of 2.7 W). It goes over the route in 200 m segments of its profile and in each one solves how fast you go with your power against the slope's gravity, rolling resistance and the air, with the air's density at the segment's altitude. No wind.

Your power. With an FTP, your chosen zone's (“Carbs you take on the bike”: Endurance 0.65 × FTP, Tempo 0.83). Without a power meter, what your usual flat-road average takes (between 10 and 50 km/h), with the same physics, at sea level and without wind; if you don't give it, you're asked for the duration: there is no assumed rider. The same power on the flat and the climbs: the starting point of pacing models (Swain, 1997; Atkinson et al., 2007), which show that pushing a little harder uphill saves time. If you do, your ride will be a little shorter and you can change the duration.

Downhill you pedal at that power up to 50 km/h and, where the slope takes you faster, you freewheel at that speed. 50 km/h is the top of an adult cyclist's typical downhill speed (32⁠–⁠50+ km/h, AASHTO guide, 2012, table 3-2); on city streets speed stops rising past −6% because riders brake (Flügel et al., 2019). No study gives a top speed: with this one an open descent comes out a little long. On a trainer there are no descents.

The typical values. The drag area (CdA) of a cyclist sitting up, hands on the tops or the hoods, is 0.36 m² (Jeukendrup and Martin, 2001, in the table of Defraeye et al., 2010); other studies range from 0.27 to 0.48 m² (Debraux et al., 2011). It is taken for 70 kg and grows with your weight to the power of 0.31, as it scales with body mass (Heil, 2001). Rolling resistance 0.004: the top of what Kyle measured for road tyres on smooth asphalt (0.0027⁠–⁠0.0040, cited by Martin et al., 1998), since real roads are rougher; Dahmen et al. (2011) validated the model on real roads with that value. Drivetrain 97.7% (Martin et al., 1998). The bike, yours if you give it in your Profile or 8 kg if not (the plan says which), plus the water of the bottles you choose.

In a sportive, in the bunch, on the flat and rolling ground the air costs you 60% of what it would alone, that of a rider second in line at half a metre (Blocken et al., 2018; in small groups, 50 to 70%); never on the climbs or descents. The plan's energy and sweat come from this same clock. The duration is always an estimate: you see it in step 02, with where it comes from, and you can change it.

References: Martin et al. 1998; Swain 1997; Atkinson et al. 2007; Flügel et al. 2019; Jeukendrup & Martin 2001; Defraeye et al. 2010; Debraux et al. 2011; Heil 2001; Dahmen et al. 2011; Blocken et al. 2018

So there is no doubt about what the calculation does and doesn't do, these lines are under study and are not applied today: the extra cost of riding off-road; the higher carb use in heat or at altitude and the higher energy cost in the cold; aerodynamic savings by terrain and position in the group; learning the power shape and NP of each ride type from your past rides; the drop in efficiency in very long efforts; a burn adjustment based on VLaMax; caffeine redosing on very long rides; sodium bicarbonate protocols; and a gut-training programme with practice rides (your ceiling does adjust with what you log in “Your ride”: “How RATIO learns your ceiling”).

At 1,500⁠–⁠2,500 m you lose a little more water through breathing (a few tens of ml per hour), within the margin of the fluid range; RATIO doesn't add it. On stays of several days at altitude, go by thirst and your weight.