The Bicing E-Bike Guide: How to Find the Best Bike Before You Unlock It

Bicing e-bikes in Barcelona are not all the same. Two bikes parked at the same station can have completely different pedal assist — one kicks in instantly, the other makes you work hard before the motor helps. This difference is well-known) among Bicing users, yet finding your preferred type without knowing what to look for usually means unlocking several bikes in a row, wasting time you did not plan to spend.

As a loyal Bicing customer and Data Science student, I became curious about whether these differences could be spotted faster — without the trial and error. So I collected data on 300+ bikes across the city, and what I found was surprising.

In this post, I’ll explain how the e-bikes work, what the data revealed, and how those findings led to a simple method to identify which type you’re getting before you unlock it.

Short on time? Skip ahead to “How to Pick the Right Bicing E-Bike: A Practical Step-by-Step Rule”.

Language: 🇪🇸 Español | 🇬🇧 English

Cadence vs Torque Sensor: Why Two Bicing E-Bikes Feel Completely Different

The answer comes down to a single component: the pedal sensor. Bicing is gradually replacing the old sensors with new ones across its fleet (a process expected to continue until the end of 2026), but the change is invisible to riders. To understand the difference, let’s look at the two systems.

The older version, which I call Type A, uses a cadence sensor design. It simply detects that the pedals are turning and activates the motor accordingly. There are several reasons why this version is being replaced:

• Poor water resistance — the sensor may cause the bike to temporarily lose assistance in rain.
• Instant acceleration — the motor engages immediately, which can lead to unexpected bursts of speed and has been linked to safety concerns.
• Shorter lifespan — it requires more frequent maintenance and replacement.

The newer Type B uses a torque sensor design, which measures how hard you are pushing on the pedals and adjusts assistance based on your effort. Type B performs better from a safety and reliability perspective, but it also requires more rider effort.

While the assistance threshold of Type B could be adjusted, riders currently face a trade-off: less effort with Type A, or greater safety and control with Type B.

Because of previous injuries to both of my knees, I cannot ride Type B comfortably for long periods. For that reason, I prefer Type A. In the following sections, I explain how to identify Type A. If you prefer Type B, simply reverse the recommendations, as Bicing’s fleet currently consists of only these two sensor types.

How I Tested 300+ Bicing E-Bikes Across Barcelona

All findings are based on direct field observation. Between January and April 2026, I examined more than 300 Bicing e-bikes across the city using only the information available to any ordinary rider.

For each bike, I recorded 15 pre-ride variables, including serial number, inventory tag, pedal feel, brake resistance, tire condition, battery level, seat height, parking orientation and more. Station elevation and capacity were later added from Barcelona’s open station dataset.

The evaluation process was strictly controlled: subjective observations were recorded before any motor-assistance testing to avoid bias. Final classification was determined through a short ride test. Bikes providing immediate assistance were labeled Type A, while those requiring stronger input were labeled Type B (Section 4.4 in paper), resulting in 188 Type A and 116 Type B bikes.

This approach has limitations. All observations were made by a single rider and relied partly on subjective judgment. To strengthen validity, key findings were reviewed with Bicing operations staff. Several interpretations — including e-bike sensors difference, pedal smoothness in recently serviced bikes, bike distribution system— were refined through discussions with Jonatan Anillo García — Maintenance Manager at Pedalem Barcelona (the company operating the Bicing system). The data generated hypotheses, while the maintenance team helped verify their plausibility.

Four Signs That Predict Your Bicing E-Bike Experience

I checked over 300 e-bikes across 15 different variables. Most turned out to be noise. But four produced results I never expected — and they changed the entire direction of the investigation.

1. The serial number tells you which sensor you’re getting.

Each bike carries two useful identifiers: a white serial number (ID) on the frame and an inventory tag located below the barcode near the rear wheel. The two are almost perfectly correlated (Spearman’s r = 0.99), meaning they generally increase together.

Serial ID (white number) & Inventory Tag (number in the center of wheel)

The inventory tag is the more precise indicator because it separates production batches more clearly. For example, the gap between old and new batches is clearly distinct in Inventory tag order (24k and 59k), while the corresponding transition in serial numbers is less obvious. The serial number, however, is easier to spot from a distance.

Older production batches — roughly those with inventory tags between 17k and 24k — are predominantly Type A. Newer batches, with tags above 59k, are mostly Type B.

One notable exception is the 19k batch. While most batches in the 17k–24k range have more than a 70% chance of being Type A, the 19k group is almost a coin flip. At the other end of the spectrum, the 63k batch has the lowest probability of being Type A, at only around 20%.

Scatter plot of Type A (Blue) and Type B (red) with Serial ID (Number) vs Inventory Tag for old batch.

Inventory Tag groups: e-bike count per group and chance of getting Type A

The relationship is not absolute, but the pattern is too consistent to be random. Most bikes retain the sensor type they were built with, while a smaller number have been upgraded during repairs, creating the exceptions seen in the data.

Probability of getting Type A and B for Inventory Tag and Serial Number (ID)

There is, however, one limitation to this approach. Bicing is gradually replacing the older cadence-sensor bikes, and by the end of 2026 the fleet is expected to consist entirely of Type B bikes. As that transition continues, production batch numbers will become less useful as a way of identifying sensor type.

2. How pedal smoothness hints at sensor type.

Spin a bike’s pedals backward and pay attention to how they feel. The strongest signal here is maintenance (Section 5.3 in paper). As Jonatan explained, every bike leaving the workshop receives preventive maintenance, including lubrication and drivetrain adjustment. These interventions make the pedals spin noticeably more smoothly. Over time, regular use gradually wears away the effects of this lubrication. As a result, smoother pedals are often a sign that the bike has been serviced recently.

However, the sensor type still has an indirect effect. Type A bikes tend to retain a smoother backward pedal feel more often than Type B bikes (this difference is statistically significant; see Section 5.3). One possible explanation is that the two systems place different loads on the drivetrain, which affects how quickly the lubricant wears over time.

Ratio of Type A and B by Pedaling Rate Score

So while pedal smoothness cannot reliably identify the sensor on its own, it slightly increases the odds of choice.

3. The battery paradox.

Type A bikes consistently showed lower battery levels than Type B, even after accounting for time of day and station altitude differences (this difference is statistically significant, see Section 5.4 in paper).

Bicing fleet performance trends by Serial ID: rolling probability of Type A (scaled to 3–4 range) and Battery Level smoothed with a rolling window of k=20. Red dotted line marks the start of new batch.

A plausible interpretation: cadence-based bikes (Type A) tend to deliver near-full assistance as soon as pedaling begins, which may cause the battery to drain faster; torque-based bikes (Type B), by contrast, provide assistance proportional to rider effort, drawing less power per kilometer.

That said, battery level is a noisy variable influenced by trip length, altitude, weather conditions, and how recently a bike has been docked. The pattern appears consistent, but field data alone cannot confirm the sensor as the direct cause. Even so, the signal is strong enough to remain relevant in the analysis.

4. Why high-altitude Bicing stations have better bikes.

Bikes parked at uphill stations felt noticeably smoother than bikes at uphill ones. That sounds interesting — until you ask why.

Pedaling Rate (left plot, dark green — high smoothness, red — low smoothness) and Battery Level (right plot, dark green — fully charged, yellow — halfway charged, red — uncharged) distribution

Bicing’s redistribution system runs on a continuous AI demand-forecasting algorithm. People naturally ride downhill and rarely haul a bike back up, so high stations keep emptying out. The redistribution trucks refill them — with bikes fresh from the workshop, chains just oiled. So “smooth bikes at high stations” really means “recently serviced bikes were driven up there.”

This also helps explain the battery pattern discussed earlier. Bikes arriving at higher stations after uphill rides tend to have lower battery levels because climbing requires more energy. As a result, two factors are acting in the same direction: sensor type and route characteristics. This makes it difficult to determine how much of the battery difference is caused by the sensor itself and how much simply reflects the journeys the bikes have taken.

The reassuring part is that the serial-number signal held consistently at every altitude — high stations and low stations told the same story about production batches. The redistribution pattern is a confound worth knowing about, but it did not poison the main result. It is a good reminder that an interesting pattern usually has a mundane explanation once you follow the logistics.

Combining the Clues: Building a 71% Accurate Identification Method

Individually, these clues are interesting. Together, they become useful.

In my dataset, about 62% of bikes were Type A. Any identification method therefore needs to outperform this baseline. Using all available variables — inventory tag, pedal smoothness, and battery level — a logistic regression model achieved 72.4% ± 6.0% accuracy (Section 5.9 in paper), confirming that these patterns contain real information about sensor type.

Later, I decided to build a rule-based model simple enough for anyone to apply in real life. Adding battery level to it produced no meaningful improvement — the pattern appears too complex to be captured by a simple rule — so it was excluded. The final method relies only on inventory tag and backward pedal feel, and still reaches 71.1% ± 5.4% accuracy.

There is one unavoidable limitation: bikes from the same production batch do not always carry the same sensor, as older units may receive newer sensors during repairs. As a result, no external method can be perfectly accurate. Most of the predictive power comes from the inventory tag, while pedal feel provides a smaller but meaningful improvement.

How to Find the Best Bicing E-Bike: A Practical Step-by-Step Rule

Here’s the whole thing — short enough to remember, detailed enough to actually use.

Step 1 — Decide what you want:

• Type A: instant, effortless push.
• Type B: natural feel, higher top speed, you do the work.

Step 2 — Read the E-bike serial number on the frame. Find your range here:

• Type A: 0-2100, 3900–7100
• Type B: 7100+

Step 3 — Spin the pedals backward to confirm. Smooth, oiled pedaling leans toward Type A and is also a sign the bike was recently serviced. If it disagrees with the serial number, trust the number. If none of the available bikes fall in your preferred serial number range, you can still improve your odds using this step.

The exact ranges describe early 2026 and will shift as Barcelona finishes the sensor swap. The method outlasts the numbers.

The Popular Rear-Wheel Trick: Why This Bicing Hack Doesn’t Actually Work

While I was testing a bike, an Italian man stopped, watched for a second, and waved me over to show his own trick: with the bike still locked in place, he lifted the rear wheel off the ground, pushed a pedal forward, and watched how freely the wheel spun. We didn’t share a language, but I got the idea instantly — a Bicing motor drives only the rear wheel, so lifting it off the ground removes the drag and lets you watch the wheel turn on its own. The speed of rotation can tell the difference between sensors.

Later I discovered that a lot of people use the same trick. But does it actually work?

I tested it properly (Section 5.7 in paper). Locked Type A and Type B wheels spun absolutely randomly, without any distinct pattern — because when a Bicing bike is docked, its motor is switched off entirely, so what you see is just plain mechanical resistance, the same for both.

Want the Full Data? Read the Complete Bicing Research Paper

If you’re the kind of reader who wants more than just the final rule, the full report includes all the details: the statistical tests behind every claim, the analysis used to separate real signals from operational noise, the methods used to isolate the contribution of each clue, the cross-validation behind the 71% accuracy figure, the comparison between rule-based and machine-learning approaches, and even the hypotheses that did not work out. Every figure mentioned in this article is included there in full, and both the data and code are openly available, so you can reproduce the entire analysis yourself.

**Read the full research paper →**

This post is based on independent field research on Barcelona’s Bicing e-bike fleet, with key findings confirmed in correspondence with Maintenance Manager at Pedalem Barcelona.

This article may also help if you searched for: How to Pick the Best Bicing E-Bike, How to choose the Best Bicing E-Bike, How to find the Fast Bicing Bike, Bicing e-bike types Barcelona, Bicing cadence sensor vs torque sensor, Bicing electric bike comparison, Bicing Barcelona guide 2026, perfect Bicing bike Barcelona, rocket Bicing bike, lemon Bicing bike, best Bicing e-bike how to find, how to always get a good Bicing bike, Bicing bike that goes fast Barcelona, Bicing e-bike data analysis Barcelona, how to read Bicing bike number, how does Bicing work Barcelona, what to check before unlocking a Bicing bike, Bicing electric bike not working, Bicing e-bike battery level how to check, how many Bicing stations in Barcelona, Bicing vs renting a bike Barcelona.