The definition, and why it is a sum rather than a formula
Heating degree days for a period are the sum, over each day, of the base temperature minus that day's mean outdoor temperature, with any negative result counted as zero. Written out, that is the whole method. A day whose mean was 45 degrees against a 65 degree base contributes 20. A day whose mean was 70 contributes nothing, not minus five, because the house does not bank heat on a warm day to spend on a cold one.
The flooring at zero is the part that makes an average temperature slightly wrong as a shortcut. Over a mixed month with warm days in it, summing day by day gives a higher degree day total than applying the month's mean, because the warm days would otherwise cancel out cold ones. If the period is uniformly cold the two agree closely, which is why this page offers both entry methods and a day list is the more accurate one.
Base 65 is a convention, not a property of your house
The 65 degree base comes from an old assumption: that internal gains from people, lights, cooking and appliances make up the last few degrees, so a house held at 70 needs no heating until it drops below 65 outside. Published degree day figures use it so that everyone is comparing the same quantity.
Your house has its own balance point and it is probably not 65. A well sealed, well insulated house with several occupants and a lot of appliance load may not call for heat until 55 outside. A leaky house held at 72 may be heating at 68. If you are comparing against published degree day data, use their base. If you are building a predictor from your own bills, you can find your own base by trying a few values and keeping the one that makes the relationship between degree days and fuel most consistent across periods.
The calibration is what makes it a prediction
On its own a degree day total describes weather. Divided into your own past fuel consumption it becomes a rate — gallons per degree day, therms per degree day, kWh per degree day — that encodes everything about your building at once: its insulation, its air leakage, its size, its equipment efficiency, its thermostat setting and its occupants. No modelled U-values or assumed infiltration rates are involved, because the house already told you the answer through the meter.
| You have | You get | Watch for |
|---|---|---|
| One delivery ticket and the degree days for that fill period | A usable rate | Delivery dates are not the same as the day the tank was last full |
| Two meter readings with dates | A better rate | Subtract the non-heating share first |
| A whole season of both | A rate you can trust across conditions | Check whether the rate drifts between mild and cold periods |
| Nothing yet | A weather description only | Start recording now; one winter of data is worth more than any model |
The base load deduction matters on any meter that serves more than heating. Gas that also runs a water heater and a range burns whether or not it is cold, and leaving that in inflates the rate and makes the prediction wrong in a way that grows with the mildness of the period. A summer month is the usual way to measure it, since almost nothing on that meter in July is weather driven.
Reading a difference between prediction and reality
Once you have a rate, the most valuable output is not the forecast. It is the residual: the gap between what the weather predicted and what the meter recorded. That gap has the weather removed from it by construction, so it cannot be explained by a cold snap, and it is therefore evidence about the house.
A large positive residual means the building or its equipment changed. Thermostat creep, an extra occupant, a window that stopped closing, insulation disturbed by other work, a damper or flue left open, or a boiler drifting out of tune all show up here. A large negative residual after you did work is the measurement of that work, and it is the only kind of before-and-after that survives the objection that last winter was colder. Anything inside about ten percent is noise, because meter read dates, wind, sun and the base temperature approximation all contribute at roughly that scale.
Where the method stops working
Degree days assume fuel use is proportional to the temperature difference at a constant conversion efficiency. A furnace or boiler is close enough to that. A heat pump is not: its capacity and its coefficient of performance both fall as the outside gets colder, so its electricity use rises faster than degree days do, and any backup resistance heat that engages on the coldest days breaks the proportionality entirely. Scaling a heat pump by degree days will underpredict cold months and overpredict mild ones, in both cases by more than the noise floor.
Wind, sun and occupancy are the other absences. A windy week at a given temperature costs more than a still one, a run of clear days on a house with south glass costs less, and a house full of people over a holiday costs less than the same house empty. None of that is in the base temperature. Where it matters, use degree days as the first-order correction it genuinely is, and put the remaining variation down as the residual rather than pretending it is not there. For the other side of the winter arithmetic, the heating fuel reserve calculator turns a load into days of supply and a refill trigger, the heating cost savings calculator prices a thermostat setback, and the heating cost guide covers what else is on the bill.
Questions people ask
What is a heating degree day?
It is one degree of coldness for one day, measured against a base temperature. If the base is 65°F and a day averaged 45°F, that day scored 20 heating degree days. Sum those over a period and you have a single number describing how much heating weather occurred, which is what makes two winters comparable. Days warmer than the base score zero rather than a negative, because a warm afternoon does not offset a cold night in a house that has to be heated on both. The unit means nothing on its own; it becomes useful the moment you divide your own fuel consumption by it.
Why is the base temperature 65 degrees?
By convention, and the convention encodes an assumption. The idea is that people, lighting, cooking and appliances supply the difference between the outdoor temperature and a comfortable indoor one for the last few degrees, so a house held near 70 does not call for heat until it drops below about 65 outside. Published degree day figures use that base so everyone is measuring the same thing. Your actual balance point depends on your insulation, air leakage, internal gains and thermostat setting, and it can be several degrees either side. Use 65 to compare with published data, and try other bases when you are fitting a rate to your own bills.
How do I find my own fuel use per degree day?
Take a period where you know exactly how much fuel was consumed — two meter readings with dates, or a tank filled at the start and filled again at the end — and calculate the degree days over the same dates with this page. Subtract anything on that meter that is not heating, such as water heating and cooking, then divide the remaining fuel by the degree days. That number is your house. Repeat it for a second period and see whether the rate holds; if it drifts a lot between a mild period and a cold one, your base temperature is probably wrong or you have a heat pump.
Does this work for a heat pump?
Poorly, and the failure is systematic rather than random. Degree days assume a constant conversion from temperature difference to fuel. A heat pump violates that in two ways at once: its capacity falls as it gets colder, so it runs longer, and its coefficient of performance falls too, so each hour of running costs more electricity. On top of that, backup resistance heat engaging on the coldest days changes the conversion abruptly. The result is that a degree day scaling will underpredict a cold month and overpredict a mild one. Fitting separate rates for mild and cold periods gets closer, but the honest answer is that the method suits combustion equipment better.
My bill went up but I changed nothing. What does this tell me?
It separates the two possible explanations, which is the main reason to bother with degree days at all. Work out the degree days for both periods and compare fuel per degree day rather than fuel. If the rate is the same, the weather did it and the house is behaving exactly as before. If the rate rose, the weather did not do it, and something on the house or equipment side changed: a thermostat setting, an occupancy change, a door or window not sealing, disturbed insulation, or equipment losing efficiency. Without the normalisation, those two very different situations look identical on the bill.